While there are elements of truth to this conventional narrative, it may also be misleading in important ways, ways that inform the most promising avenues toward overcoming our political paralysis. First, there is significant evidence that, for most people, spending time on the Internet and on social media increases rather than decreases their exposure to ideologically diverse content (in addition to increasing exposure to opinion-reinforcing content).1 1. Pablo Barberá, “Social Media, Echo Chambers, and Political Polarization,” in Social Media and Democracy 34, 37–40 (Nathaniel Persily & Joshua A. Tucker eds., 2020); Seth Flaxman, Sharad Goel, & Justin M. Rao, “Filter Bubbles, Echo chambers, and Online News Consumption,” 80 Public Opinion Quarterly (2016), 298; Matthew Barnidge, Exposure to Political Disagreement in Social Media Versus Face-to-Face and Anonymous Online Settings, 34 Political Communication (2017), 302; Laura Silver, Christine Huang, & Kyle Taylor, “In Emerging Economies, Smartphone and Social Media Users Have Broader Social Networks,” Pew Research Center (2019).Part of this counterintuitive phenomenon might be explained by the fact that inadvertent exposure to diverse content is more common online, but it might also reflect the fact that a significant proportion of the people we engage with online are not embedded in our closest social networks, and therefore are not (in a sense) prescreened for ideological conformity.2 2. Barberá at 41.

Second, there is also evidence that exposure to diverse viewpoints increases rather than decreases mutual distrust and political polarization.3 3. Barberá at 46–47; Jaime E. Settle, Frenemies: How Social Media Polarizes America (2018). How might this happen? There may be good reason to think exposure to contrary ideas affirms our political identities and feeds into so-called “affective” polarization, which is grounded less in specific issue positions than in partisan social identity.4 4. Barberá, supra note 3. Awareness of the views of others can reinforce all the ways they are different from us and can inhibit our ability to be persuaded by their perspective.5 5. Lilliana Mason, Uncivil Agreement: How Politics Became Our Identity (2018).

One need not take a definitive position on this research, which is mixed, to recognize that any approach to diversifying the information ecosystem for individual users must take account, not just of the bare existence of diverse information flows, but also of its actual effects on the behavior of listeners and readers. Understanding which interventions are most productive also requires assessing what, precisely, exposure to information is meant to achieve.

Consider four different justifications for exposing people to diverse information:

• Understanding. We might expose people to diverse information to promote understanding and awareness. Someone who receives information from a narrow range of sources may not be receiving a full factual picture and may not be hearing the range of arguments in support of or against a policy.
• Persuasion. We might also seek to expose people to diverse viewpoints to try to persuade them that those viewpoints are correct or should be adopted.
• Self-fulfillment. We might wish to promote diversity of viewpoint so that people holding diverse views obtain the fulfillment of expressing themselves and being heard. This rationale could also be expressed in terms of equality of opportunity for diverse voices to influence the public sphere.
• Collaboration. We might make people aware of diverse information so that information can serve as an input into decision-making. This rationale is distinct from “understanding” because it is action-oriented, and it is distinct from “persuasion” because the interest in collaboration can be satisfied without the information actually convincing anyone to change their view.

These different justifications for diversifying someone’s information environment support different kinds of interventions. For example, bringing diverse speakers onto a college campus or promoting the visibility of diverse content on social media may increase awareness, but inducing people to be open to changing their views may require attention to the speaker’s identity and their rhetorical strategy. Failing to be attentive to these dynamics may promote awareness, self-fulfillment, or equality of opportunity—the values that not incidentally are heralded as underlying the First Amendment’s commitment to robust free expression—but it can be counterproductive in sustaining social trust, much less increasing it.

To my mind, the public discourse has not paid sufficient attention to the fourth of these justifications for exposure to diverse information: enabling collaboration. A personal anecdote may help set the stage. In 2022, the National Constitution Center convened three groups of scholars from diverse political perspectives—a “progressive” group, a “conservative” group, and a “libertarian” group. (I was a member of the first group.) In the first stage of the project, each group drafted its own ideal constitution. In the second stage, the groups convened a virtual constitutional convention and drafted amendments to the Constitution. Unsurprisingly, each of the “ideal” constitutions drafted in the first stage, among fellow ideological travelers, was quite different from the others. Remarkably, though, the groups were able to agree to five amendments that, if adopted, would be fairly consequential.

The key to achieving this level of consensus wasn’t that there was substantial agreement among the participants—indeed, the participants were truly ideologically diverse. The first stage of the project made each group aware of the positions of the others, but speaking for myself, this awareness was at least as alienating as it was unifying, as it clarified the degree of difference between the groups. Moreover—again speaking for myself—learning that some particular reform was supported by other participants led me to question rather than affirm my own support, as it made me worry that I had not sufficiently reflected on the consequences of the reform. This is how affective polarization works, and it’s a heck of a drug.

Nor did consensus over amendments follow from successful persuasion among the groups. The second stage of the project did not, so far as I can recall, cause anyone to revisit their highly disparate positions as to what an ideal constitution should contain.

What the collaboration stage did instead was force the groups to take collective action. Doing so requires modifying one’s position to accommodate the positions of others. Importantly, collaborative action does not require agreement or persuasion. It does require a shared commitment to the collective enterprise. Understanding the beliefs and commitments of other members of the collective enabled us to propose informed compromises.

My own experience with the National Constitution Center is consistent with longstanding research in social psychology dating back at least to Muzafer Sherif’s foundational “Robbers Cave” experiment in the 1950s.6 6. Muzafer Sherif, Intergroup Conflict and Cooperation: The Robbers Cave Experiment (1961). This research suggests that mere contact or coexistence can be unhelpful in mitigating intergroup conflict, but that requiring conflicted groups to focus on what psychologists call a “superordinate” task—one that no single group can accomplish alone—can induce cooperation and improve intergroup relationships.7 7. Kelly Kirkland et al., Promoting Prosocial Behavior in an Unequal World, 13 Frontiers in Psychology (2023).

My little anecdote and this body of research do not offer a panacea. Even on its own narrow terms—a fake constitutional convention—participants in the constitutional drafting project were not especially diverse along professional or other socioeconomic dimensions, and there is evidence that economic inequality can impede intergroup cooperation.8 8. Kelly Kirkland, Jolanda Jetten, & Mark Nielsen, The Effect of Economic Inequality on Young Children’s Prosocial Decision-Making, 38 Developmental Psychology 512 (2020). Still, they serve to emphasize that shaping the flow of information depends in large part on what one hopes information will accomplish, and that collaboration is an important objective we may wish to prioritize.

We would do well then—or better, anyway—to imagine ways in which we don’t merely communicate, but also are required to collaborate, across difference. This might mean increasing funding for deliberative mini-publics, what the political scientist Robert Dahl called “minipopuli,”9 9. Robert A. Dahl, Democracy and Its Critics 340 (1989). tasked with offering targeted governance proposals. Universities could increase funding for programming jointly designed by ideologically diverse groups, such as the Federalist Society and the American Constitution Society. Media outlets could be induced to provide dedicated space for co-authored commentary that provides specific policy recommendations. Platform algorithms could boost content cross-posted across ideological difference. We could continue to disagree, as we always will, but we could seek to innovate around disagreeing constructively, rather than merely for its own sake.

“It’s disappointing that the Court failed to take this opportunity to make clear that public servants can go directly to court to challenge broad restrictions on their speech,” said Alex Abdo, litigation director at the Knight First Amendment Institute. “Forcing public employees to wade through cumbersome and potentially futile administrative proceedings before challenging prior restraints allows unconstitutional censorship to persist. Now more than ever, we need the insights of the nation’s immigration judges and other public employees to understand the work of our government.”

In 2020, the Knight Institute filed this lawsuit on behalf of NAIJ, challenging an Executive Office for Immigration Review policy that prohibits immigration judges from speaking publicly in their personal capacities about immigration or the agency that employs them. A district court dismissed the case in 2023, holding that the Civil Service Reform Act of 1978 requires such claims to proceed through administrative review.

In June 2025, the Fourth Circuit revived the case, expressing concern that recent actions by the Trump administration may have undermined the agencies charged with hearing federal employment claims—the Office of Special Counsel and Merit Systems Protection Board—and that they may no longer be operating as Congress intended. But the court also held that, if that system is functioning as Congress intended, immigration judges would be required to challenge the policy through that administrative process rather than in federal court.

In February, the Knight Institute filed a cross-petition for certiorari asking the Court to consider that question. The cross-petition was filed after the government petitioned the Court in December 2025, asking it to reverse the Fourth Circuit’s ruling. 

“We’re disappointed by today’s decision and by the Court’s failure to address the significant free speech concerns at the heart of this case. However, this litigation is far from over,” said Holly A. D’Andrea, president of the National Association of Immigration Judges. “The case has been remanded to the Fourth Circuit, and NAIJ will continue fighting to protect the free speech rights of immigration judges, to seek meaningful review of the Executive Office for Immigration Review’s speech policies, and to ensure that immigration judges may engage in public discourse on immigration matters in their personal capacities. Justice cannot endure when judges are intimidated into silence, nor can a nation remain free when the rule of law is subordinate to the whims of political ambition.”

NAIJ is a nonpartisan, nonprofit voluntary association of federal immigration judges. Its members have long participated in public conversations and events about immigration law, including through teaching, training, and community engagement.

Read today’s decision here.

Read more about the lawsuit, Margolin v. National Association of Immigration Judges, here.

Lawyers on the case include Ramya Krishnan, Alex Abdo, Xiangnong (George) Wang, and Jameel Jaffer of the Knight First Amendment Institute, and Victor M. Glasberg of Victor M. Glasberg & Associates.  

The district court ruled in favor of WhatsApp in late 2024. A jury awarded damages in 2025, and the district court entered a permanent injunction. NSO Group subsequently appealed the court’s rulings on liability, injunctive relief, and damages.

The Knight Institute’s amicus brief describes the threat to free expression and press freedom that results from the use of commercial spyware like Pegasus. It argues that the global proliferation of spyware is not inevitable, and that U.S. courts can protect Americans and others from the harms of spyware. As the brief explains, companies involved in developing and deploying spyware to access devices without authorization can be held liable under the CFAA and California law.

Status: Briefing ongoing.

Case information: WhatsApp LLC v. NSO Group Technologies Ltd., No. 25-7380 (9th Cir.)

But when it comes to AI risks, he is far more bearish. He points to examples of cyber- and bio-risks and expresses pessimism about AI quickly becoming dangerous across many new domains. He argues that AI’s emergent capabilities make it fundamentally different from previous technologies, and that this difference justifies “extraordinary” government responses including restrictions on what companies can release.1 1. Thompson writes: "I can understand a plan to treat AI as a ‘normal’ technology and let Nvidia export powerful chips to China. And I can understand a plan to treat AI as an ‘abnormal’ technology that compels the government to create extraordinary regulations that prevent private companies from selling their products and services on the grounds that they’re too dangerous" [emphasis ours]. He goes on to conclude that AI is, in fact, abnormal, implying support for extraordinary government intervention. Our essay is a response to that conclusion.

In this essay, we lay out the downsides of extraordinary government intervention in response to new technology. We discuss proposals for improving resilience that do not require such intervention. We also discuss why governments have so far been reluctant to invest in resilience. In short, resilience requires us to get better at the *normal* process of policymaking. But sclerosis in the federal government and the ease of justifying interventions on AI companies rather than society at large make extraordinary intervention seem appealing, despite its limitations.

What happens next? Policymakers can either get their act together and invest heavily in improving resilience, or be forced to take extraordinary actions (such as on AI nonproliferation) that are more onerous and less effective.

Nonproliferation is brittle because it relies on a single chokepoint. Resilience distributes defenses across society.

A Recap of AINT’s Arguments and Thompson’s Areas of Agreement and Disagreement

Many people, including Thompson, have found AI as Normal Technology to be a useful framework for thinking about AI’s economic impacts while being unconvinced by our views on safety. In the AINT essay, we make different arguments about slow labor market impacts and resilience to misuse risks.

The labor market argument rests on the speed of diffusion: there are many speed limits between a new AI capability and its economic impact, including the need to build products, change organizational workflows, and navigate regulation. This has proven helpful both to understand why claims of rapid and widespread job displacement (such as Amodei’s claims about an imminent white-collar bloodbath) are unlikely to materialize, as well as to identify bottlenecks to diffusion that could hinder the beneficial adoption of AI.

But our argument about AI’s misuse risks depends on the offense-defense balance: whether attackers or defenders benefit more from a given capability, and our ability to build resilience in response to misuse risks. Notably, an attacker does not need to go through the slow process of organizational adoption to cause harm. We argue for building societal resilience to intervene on misuse risks and identify many interventions that we should take to reduce AI risks.

Here, Thompson disagrees with us. One reason he disagrees is that all general-purpose technologies bring about new risks, which are hard to reason about from previous patterns. But even in the history of general-purpose technologies, Thompson argues that AI seems particularly “abnormal” because of risks that are emergent and unknown even to AI developers. And this justifies plans by the government to treat AI as an abnormal technology, since it “compels the government to create extraordinary regulations that prevent private companies from selling their products and services on the grounds that they’re too dangerous.”

Thompson doesn’t define precisely what he means by extraordinary government action, and the specific interventions he envisions remain somewhat vague. Since we’ll be arguing against such interventions, we want to be clear about what we mean by the term. We think of extraordinary interventions as a spectrum defined by three factors that apply to any powerful technology, not just AI.

First, extraordinary interventions tend to be precautionary. They restrict activity based on anticipated harms rather than realized or demonstrated ones. This doesn’t mean that precaution is never warranted, but it does mean the justification needs to be stronger, since we are restricting activity without clear evidence that the harms will materialize (or that they’ll materialize exactly as we predict). Even then, an intervention is more extraordinary when viable alternatives exist that can address the risks while being less restrictive, such as investing in resilience.

Second, extraordinary interventions impose restrictions on the liberty of actors who are not directly responsible for the harms in question. When governments restrict what AI companies can release, the burden falls not on the malicious actors who cause harm, but on companies that build tools that could, in principle, be misused. This is especially pertinent for dual-use technologies. Because these tools have widespread beneficial applications, restrictions on companies can cut off beneficial access for the broader public in order to prevent misuse by a small number of bad actors.

Third, extraordinary interventions bypass normal processes of governance, and instead rely on unilateral authority such as emergency declarations or executive orders, even though the governance processes being bypassed exist to ensure that restrictions on liberty are subject to democratic accountability.

An intervention need not satisfy all three of these criteria to count as extraordinary. But the more of these factors that are present, the higher the bar should be for justifying it.

Extraordinary Government Action is Extraordinarily Costly

We agree that AI poses misuse risks. But our experience from regulating “abnormal” technologies shows how burdensome extraordinary government interventions can be. The enforcement of nuclear nonproliferation has required the IAEA, the Non-Proliferation Treaty, decades of diplomacy, ongoing investments, and even military confrontation. The tolls of these interventions were enormous, but the approach was at least somewhat enforceable because nuclear weapons depend on enriched uranium, a physical bottleneck that is genuinely hard to get around.

AI is different from nuclear weapons. For one, there is no equivalent “physical” bottleneck. The core techniques for building AI systems are well known. Adversaries (especially nation-states) can match frontier capabilities within months. Any nonproliferation regime for AI would face constant erosion.

In the face of this challenge, how could governments maintain nonproliferation? Some proposed interventions, such as the rumored executive order on voluntary commitments by AI companies for predeployment evaluations, are relatively low on the scale of extraordinary interventions. Done well, they could allow us to tilt the offense-defense balance by giving defenders more time to prepare for new capabilities.

The U.S. has also enacted export controls on chips. We think this is a modest intervention: countries routinely restrict exports of sensitive goods to maintain their lead in innovation, and reasonable people can disagree about where to draw the line. But when it comes to preventing dangerous capabilities from being widely accessible, export controls are far less effective. Open-weight models and widespread API access from frontier labs mean that the gap between frontier and publicly available capabilities is at most a few months, not years.

If the most that nonproliferation can buy us is a few months, the urgent priority should be investing in resilience so that we are better prepared when those capabilities inevitably become widely available, as we discuss below. On the other hand, if governments try to maintain nonproliferation as a way to deter the availability of advanced AI capabilities for misuse, their interventions will necessarily get more demanding.

To truly enforce nonproliferation, governments would need to restrict access to open-weight models and even API access to capable models. This would require licensing regimes that give governments ongoing authority over which models can be released and restrictions on open-weight models. We might quickly enter a state where governments exercise control over what AI research and products can be shared publicly. In fact, we have already seen concerning examples of such control, such as Anthropic’s designation as a supply chain risk, and recent rumors about licensing requirements for AI companies.

To avoid this slippery slope, those proposing extraordinary interventions should be clear about where they would draw the line. Would they support restrictions on open-weight models? What about requiring approvals for each new model release, or restrictions on the movement of researchers who build frontier AI across countries? If proponents cannot specify the limits of what they are calling for, it is reasonable to expect that the demands for increases in the government’s ability to take unilateral action will keep escalating as capabilities advance.2 2. Thompson's essay conveys support for restrictions on AI companies but remains vague about what those restrictions would look like in practice. Our analysis of what interventions would be necessary, and their downsides, is informed by our own thinking on the topic rather than his specific proposals.

For nuclear weapons, once the nonproliferation regime was built, it did not need to be rebuilt every few years. But AI is not the last digital technology with powerful dual-use properties. New technologies would raise similar questions and potentially demand similar or escalating responses. So the “abnormal technology” framework for regulating AI would start to look less like a targeted response to a specific risk and more like a permanent expansion of government powers over what citizens and companies can build, publish, and research.

We have seen this debate play out before. Each new technology raises questions of what restrictions to liberty are appropriate to mitigate harms. This history tells us we shouldn’t automatically default to imposing extraordinary government interventions.

The internet allowed people to access information about how to build bombs, and in 1995, after the Oklahoma City bombing, the late Senator Feinstein introduced a bill that, in its first draft, would have criminalized the distribution of any information on the internet describing bomb-making materials or processes. (When the bill was eventually passed, it had a narrower scope, requiring offenders to have knowingly aided a crime through their instructions.)

The federal government also tried to restrict access to encryption software, arguing it would help criminals communicate beyond the reach of law enforcement. It imposed export controls and proposed requiring backdoors in encryption so that the government could always access private communications. It even started a criminal investigation against a programmer under the Arms Export Control Act for releasing encryption software.

A t-shirt with the source code for the RSA encryption algorithm, used to protest the government’s restrictions on cryptography. Source

These restrictions were eventually rolled back through a combination of court rulings and executive action. Encryption became the foundation of digital security, enabling e-commerce, online banking, and many other applications.

Yet, at other points in history, we have accepted increases in government powers in response to new technologies. The widespread acts of terrorism after the invention of dynamite catalyzed the expanding government surveillance apparatus by the FBI.

The question is whether this level of intervention is necessary to address AI risks. One response that does not require extraordinary government intervention is improving resilience.

Resilience is an Under-Emphasized Defense Against AI Risks

Resilience is the capacity of a system to withstand and adapt to harm. It was one of the main defenses we proposed in the AINT essay. Unlike extraordinary government actions like restrictions on what companies can release, resilience does not impose costs on AI companies. Instead, it focuses on improving our capacity to respond to and recover from AI risks, regardless of when or where they occur. For past technological harms, improving the resilience of systems has been key to reducing harm.

Consider cybersecurity. The internet created entirely new classes of attacks, such as worms that spread through networks, causing billions of dollars in damages. The way we defended against these risks was not by restricting access to computers or the internet, but rather through bug bounties to incentivize people to report vulnerabilities to developers, improving browsers and operating systems, automated testing, and better patching practices. All of these improved the resilience of our cyber infrastructure regardless of where the risks arose.

Automated vulnerability detection tools are another example of technology enabling new kinds of attacks. Tools such as fuzzers and symbolic execution engines have been “superhuman” at vulnerability detection for years; they can detect vulnerabilities at a scale humans cannot match. Yet, they have been freely available on open-source repositories. Since defenders had access to the same tools, they became core defensive tools, largely funded by the cyberdefense ecosystem. In fact, defenders had structural advantages in using these tools effectively, such as deeper access to the systems being tested. This in turn led to better protections for cybersystems.

Stylized spectrum of vulnerability detection capability, showing that existing tools were already vastly superior to unaided vulnerability researchers.

This shows how we have managed the transition from unaided vulnerability detection to vastly superhuman detection without requiring extraordinary government intervention. We agree that LLMs bring real improvements to vulnerability detection, but they build on top of decades of tooling that is already far beyond what any human can do unaided. If we have already absorbed the transition to such tools, it is worth asking whether the additional capabilities of language models call for extraordinary interventions.

To be clear, these transitions were not smooth or painless. For a brief period, the attacker-defender balance was completely upset. The idea of viral spread of malware and the resulting asymmetry was unprecedented. Fifteen-year-olds could create devastating cyberattacks that brought down top e-commerce websites and search engines. It is plausible that government intervention to improve cybersecurity could have reduced the harm caused to individuals and businesses during this period.

AI’s use for cyberattacks might once again upset the offense-defense balance, and we don’t think the AI transition will be smooth by default. Many systems that are currently under-defended, including schools, hospitals, power grids, and small government agencies, will be at real risk. Efforts like Project Glasswing and OpenAI’s cybersecurity grant program are important but insufficient on their own. While restricting access to advanced AI systems might be helpful in the short run, it is not a silver bullet in a world where open-weight models are only months behind the most capable closed models.

Addressing AI’s cyberrisks requires investment in resilience. That means AI-assisted red-teaming not just for tech companies, but for schools, hospitals, power grids, small businesses, and government systems that currently lack the capacity for defense. We should also incentivize professional security experts to find and report vulnerabilities (with or without AI use), such as through bug bounties that cover more than just tech company products. These efforts are not automatic; they require investment and planning.

The good news is this work is starting across many different industries. While this requires significant investment, it is still far less burdensome than enforcing a strict nonproliferation regime. The bad news is governments have a lot left to do to truly make the transition painless.

The same analysis applies to biosecurity. AI may lower some information barriers, but bioattacks depend on many downstream steps: procuring materials, accessing specialized equipment, and applying tacit know-how. We can intervene on those downstream steps *now* to increase societal resilience to bioattacks without imposing strong controls on AI development, such as by implementing better screening of synthetic biology orders, using AI to evaluate the riskiness of new compounds, tracking access to dangerous materials, and offensive red-teaming (such as asking trusted experts to attempt to use AI for manufacturing dangerous compounds) to find gaps in these efforts. These help regardless of whether biorisks are from AI.

If Resilience is So Helpful, Why Haven't We Prioritized it Already?

Importantly, resilience does not require extraordinary government intervention. It only requires us to get our act together on the “normal” process of policymaking and execution.

The problem is we are not great at normal policymaking. We suspect one reason why the resilience approach seems unappealing is that it requires polycentric governance in which many decision-makers work harmoniously together. This is a tough sell given that state capacity in the United States has been hobbled by decades of accumulating veto points and creeping proceduralism. As a result, unilateral actions by the executive branch are often seen as the way out for developing and enforcing AI policy.

Investing in resilience requires action and investment by a much wider set of actors than nonproliferation does. For resilience to be effective, the government needs to legislate and allocate funding, collaborate across agencies, build early warning systems, and serve as a resource hub for downstream actors and rapidly disseminate information to help them shore up their defenses. The U.S. federal government isn’t exactly well known for being competent at this set of tasks.

So it shouldn’t be a surprise that when we look at the government’s track record, policy responses for resilience have been underwhelming. In fact, even when there are areas of agreement between proponents of the normal and “abnormal” views on AI, such as requirements for transparency, auditing, and safe harbors for safety research, we are yet to see concrete federal action. (There are some efforts, like the recent letter from representatives urging the federal government to improve coordination on cyberrisks. But for now, it is just a letter with recommendations; it remains to be seen if any action will be taken based on these suggestions.)

It is in this context that extraordinary government actions, such as nonproliferation, look tempting to address AI risks. These are morally satisfying since they primarily impose burdens on the companies that create these risks. They are also tractable, since they only require unilateral action from the executive, such as invoking the Defense Production Act. The main question is whether we can achieve better outcomes through such actions, or whether we should invest in improving the “normal” process of governance.

While we understand the reasons for taking the former approach, we lean towards the latter. Getting our policy act together is hard, but important—not just for this round of AI policy and misuse risks, but for all future interventions on technological harm, and for the democratic process to work more generally.

On the other hand, if our main defense against AI risks is nonproliferation, a single technical breakthrough that makes models cheaper to train could be enough to cause instability, especially in a world where we don’t also invest in resilience. The damage will be that much greater when the dam eventually breaks.

Acknowledgments

We are grateful to Katy Glenn-Bass for feedback on a draft of the essay and Shira Minsk for editorial support. 

Today’s “Media Individualism Complex” has five powerful dimensions that feed each other: (1) celebrity influencers with personal brands and loyal audiences who use their embodiment to command attention; (2) billionaires owners and celebrity CEOs using media companies for economic gain, personal advancement, and political positioning; (3) critics, intellectuals, and branded academics invested in styles of commentary and critique that need and feed individual reputations; (4) media technologies like personalized platforms and chatbots that watch, aggregate, model, and feed individual actions; and (5) news coverage that purposefully blends media figures’ work and personae.

Distinct from media systems driven by a concern for public goods, collective action, and procedural accountability, the Media Individualism Complex has its own logics. It is shaped by embodied, personal interests and private wealth, and also thrives on and strategically invokes notions of benevolence, fame, shame, charisma, reputation, celebrity, propriety, exclusivity, courage, genius, and other individually possessed and performed forms of power. The Media Individualism Complex may periodically intersect with and even support public interest media (celebrities supporting causes, philanthropists endowing a newsroom, patrons funding a documentary, star academics advocating for social justice), but the complex is fundamentally at odds with public interests because its people, signals, norms, and ideals systematically and structurally traffic in individualism.

To be sure, this complex is not new; individuality and personality have always been at the heart of media systems, especially in the United States. The country’s media has long been dominated by celebrity influencers (Charles Coughlin, Walter Cronkite), owner-patrons (William Randolph Hearst, Walter Annenberg), branded critics (Walter Lippmann, Gore Vidal), individualism publications (People Magazine, TV Guide), and coverage that mixes biography, personae, and public interests (Woodward and Bernstein portrayals in “All the President’s Men,” and the embodied performative interview styles of brands like Barbara Walters, Kara Swisher, and Ezra Klein). Such individualism is not to be dismissed or derided. It has sometimes made laudable and powerful interventions into media systems—e.g., when Newton Minow delivered his “vast wasteland” speech, when Jon Stewart accused Tucker Carlson of “hurting America,” and when James Baldwin mixed memoir and critique to denounce Hollywood racism.

Today, this Media Individualism Complex plays out amongst:

• Celebrity Influencers. Media figures like Charlie Kirk, Joe Rogan, Hasan Piker, Stephen Colbert, Jimmy Kimmel, and more regularly mix celebrity, personal politics, humor, and populism in ways that shape the Media Individualism Complex messages and styles. Sometimes canceled and even killed for such individualism, they show the power and precarity of starring in a system that prizes embodied, solipsistic analyses. And the rapid rise of hybrid influencer-journalist cultures means that social media stars with personal brands and followings increasingly drive news coverage. Much rides on what these individuals say, the risks they take, and how well they leverage funders, broadcasters, politicians, audiences, platforms, and attention economies.
• Billionaires & Funders. Wealth and power has also rapidly concentrated in a relatively small set of individuals who have become key actors in creating, funding, and running the Media Individualism Complex. Billionaires like Jeff Bezos, Mark Zuckerberg, Elon Musk, and Patrick Soon-Shiong own and shape media and technology companies according to their personal values. CEOs like Disney’s Bob Iger (Disney owns ABC News) and Paramount’s David Ellison (Paramount owns CBS News) have been at the center of politically charged questions about how their media companies will reward, censor, or allow speech, and how such decisions weigh on their personal legacies and political influence. And wealthy and famous individuals like Dave Chappelle, Joan Kroc, Connie Ballmer, Eric Schmidt, Pierre Omidyar, and Craig Newmark all choose which parts of the media system to endow. And entire literary traditions focus on how media systems make and sustain such personalities.
• Critics & Intellectuals. Many critics have built careers on critiquing concentrations of power in ways that both need and feed the Media Individualism Complex. Celebrity intellectuals with publicists, agents, assistants, stylists, and named chairs maintain powerful personae through a system of columns, podcasts, TED talks, exhibitions, social media, magazine profiles, “genius” prizes, fellowships, lists curated by Forbes, Time Magazine, MIT Tech Review, and Business Insider, and with personal access to elite gatherings like Dialog, Bilderberg, and the World Economic Forum. While such individuals may use their celebrity to challenge media power in ways that strategically align with or are benevolent to public interests, their ability to do so requires a system that values and trades on individualism.
• Platforms & Designs. As much scholarship has documented, the algorithmic logics of social media platforms are often driven by analyses of how well content engages people as individuals—whether they are entertained or outraged, whether they scroll past or repeatedly view, whether they comment or share, whether they become habitual users. Platforms sense personal reactions, prompt individuals to act, and try to elicit embodied emotions that people are encouraged to represent and share with others. A dogged focus on individual engagement is crucial to ensuring that people repeatedly return to the Media Individualism Complex and pay attention to its speakers, funders, critics, and coverage. This automated individualism is especially pronounced in contemporary AI bots designed to be agreeable, confirm individuals’ impressions, and foster a kind of sycophantic intimacy.
• Persona Coverage. Even coverage by media organizations reinforces the power of media individualism, with publications regularly producing lifestyle profiles of celebrities, influencers, billionaires, and critics. The Guardian ponders the significance of a Mark Zuckerberg AI avatar, and the New York Times asks if Tim Cook was a stylish CEO, speculates on the influence a “MAGA Girlfriend” has on Google co-founder Sergey Brin, analyzes the Sam Altman-Elon Musk feud, and publishes a profile of CEOs asking whether CEOs should be profiled. The Wall Street Journal writes more than 800 words on how a sweater emblazoned with a CEO likeness shows how “leaders at companies from Nvidia to Palantir are now driving fashion, signaling a new era of the cult of the founder.” Vogue magazine devotes almost an entire issue to Jeff Bezos’s wedding. The Media Individualism Complex needs such figures to be personalities worthy of coverage, recursively reinforcing its own power and the power of those they cover by making audiences see celebrity as unquestionably, naturally newsworthy.

Though charm, charisma, and celebrity can be leveraged for ill effects, individuals can also harness them for good. They can use their individuality to call for changes that they think will make the media systems better serve public interests. To be clear, I do not decry all forms of media individualism but it is incredibly dangerous for media systems to not ask why certain individuals are so powerful, how individualism has become such a powerful media ideology, and what happens to media systems when reformers must repeatedly fit public interests into individualized, private interests, benevolences, and largess of those who populate, make, fund, and police the Media Individualism Complex.

What might be done? There are a few ways forward.

First, individuals may retain their personal power, but use it to work against the very forces that gave it to them, the forces that sustain the Media Individualism Complex. They could see themselves as embodiments of failed policies and systems and work to make themselves the last exemplars of such failure. Second, individuals could ‘quit’ their positions as powerful media individuals and find ways to ‘donate’ their personal power to collective causes, divesting from positions of individual privilege in ways that reassign individualized power to publics. Third, individuals might organize collectives that pool their personal power into collective instruments of influence that they direct through shared governance and collective action. They would create new forms of power that were fundamentally anti-individual. Fourth, individuals may simply become more honest and transparent about having and using private power, neither giving it up nor reassigning such embodied privilege, but being explicit about why they think they should have and use it.

Fifth, powerful individuals may better articulate how their personal power rests upon people or forces that are largely invisible but that are nonetheless essential to creating and sustaining their seemingly independent power. They might call attention to—and make themselves responsible to—the apparatuses and infrastructures that make them seem special and singular. Sixth, though it seems to me like an ethical race-to-the-bottom, some people may adopt a “ridicule as praxis” approach that personally shames, demeans, or otherwise demotes powerful media individuals, centering and calling attention to their personal absurdities. Seventh, media organizations could use their coverage to critique the fundamentally unacceptable idea that some individuals should have more or less media power, highlighting individualism as dangerous and antithetical to public interests, not an idiosyncratic or quirky curiosity designed to entertain. Finally, academics could follow in the footsteps of scholars who have studied charm, charisma, celebrity, and mogul power to develop precise and critical accounts of how media individualism works and shapes media power, and how billionaire media figures might be countered.

If media systems are to ensure free expression they should challenge increasingly personalized concentrations of wealth, status, and technological control. Concentrated power needs people to think that media systems dominated by individuals are inevitable. They are not; things could be otherwise. At best, powerful people are benevolent individuals who can be convinced to invest their private power in public interests; at worst, these are self-interested personalities who see their power as a natural, meritocratic reward that entitles them to impose their own ideas of media on those who struggle to tell and listen to stories without simply being clients of the interests of wealthy and powerful individuals.

Finally, holding a mirror to ourselves and those close to us, we should also see how critics, reformers, and those who have suffered the ills of a broken media system often take up individualism as a tool of progress. They understandably think that their own benevolent applications of individualism—serving personal theories of structural change and social justice—are ways out of the brokenness. They see their ‘correct’ application of individualism as a form of resistance. They know how precarious their individual power is, and they know the private compromises that they have had to make to retain it. They may choose to invest in and celebrate the very forces of individualism that they know work against the long term, collective, public projects that they know emerge from and sustain a robust media system. Such compromises are understandable but dangerous because they ultimately sustain the Media Individualism Complex. It is paradoxical, but the most courageous thing that a powerful member of the Media Individualism Complex might do is actively work against the forces that sustain their power, and that make their courage relatively safe and easy.

The U.S. government, however, has remained a global outlier for how little it funds its public media. Even before the 2025 rescission package nullified federal support, U.S. annual expenditures were approximately $1.60 per capita—almost literally off the chart compared to most leading democracies. To even call our system “public” is a misnomer; it has long been funded primarily by private sources of capital from individuals, foundations, and corporations.

This structural contradiction—public media funded by private capital—has rendered the U.S. system politically and economically weak. Even so, public media institutions have often provided higher-quality fare compared to their commercial counterparts. This became especially notable during the Trump 2.0 era when privately-owned, commercial media have capitulated to various forms of oligarchic and authoritarian capture.

For these reasons and more, we must establish independent public media if we’re to achieve any semblance of U.S. democracy. Moreover, the current media poly-crisis offers a rare opportunity to entirely reimagine and rebuild our public media system. The following analysis begins to sketch out the structural conditions necessary for creating a new public media system anchored by what I term the “Public Media Center.” Broadly speaking, three organizing principles—de-privatization, democratization, and localization—can help guide us toward reconstructing our public media from the ground up.

De-privatization

From its earliest days, the American public media system has depended heavily on private philanthropy and, increasingly, corporate funding. Paltry government support and subsequent policy changes further incentivized public media institutions to seek out corporate sponsorships in the form of seemingly full-blown commercial ads (euphemistically referred to as “enhanced underwriting”). Such reliance on corporate money not only blurs important distinctions between commercial and public media, but it also can skew programming in anti-democratic ways. The first step toward creating a truly public media system requires sufficient, permanent, and insulated funding to support high-quality and universally accessible media. Ballpark calculations based on international standards and historical precedents put this budget at roughly 30 billion dollars per year. This money could be allocated via block grants to states and further devolved though public grant-making like those deployed by civic information consortiums aimed at supporting local journalism.

Democratization

A growing body of research demonstrates positive correlation between independent public media systems and healthy democracies. However, public media institutions themselves also should be democratized. To do so isn’t only ethically sound, but also strategic: Democratization can help ensure bipartisan trust, structural resilience, and journalistic independence. Toward this aim, a public media system shouldn’t be public in name only but actually owned and controlled by local communities. This collective ownership and governance must translate to public participation in all stages of media production and dissemination to encourage trust and community investment. In fact, public media institutions are often cross-ideological islands of trust in an otherwise dismal landscape. This trust can be further deepened through direct public ownership and participation, perhaps through community action programs, participatory budgeting, or some similar bottom-up model of governance. Rotating oversight boards could be locally elected or randomly selected to ensure public accountability.

Localization

Although locally rooted, public media outlets often amplify national instead of community programming. By becoming more localized in their coverage, public media outlets would realign their programming with the types of news, cultural fare, and information currently being under-provided in many communities. This refocusing also would return public media institutions to their historical roots. Prior to the Public Broadcasting Act of 1967, educational broadcasters were typically situated in community-facing stations based at public universities. But once the public broadcasting network became consolidated and centralized, it shifted toward national programs instead of showcasing local culture and public affairs—partly because producing local media is costlier than syndicating national content. While some recent initiatives focus more on providing the local journalism that the contracting newspaper industry no longer produces, public media outlets have long remained an under-utilized infrastructure in combating the ever worsening news deserts problem. To prevent undue federal control over the allocation process, public media funding could be devolved to the state and county levels—even municipalized—to better respond to community information and communication needs.

Imagining the Public Media Center

Supporting strong public media at the center of our broader information ecosystem can help guarantee a baseline level of reliable news and communication for all members of society—a universal service mission that commercial systems rarely provide. Moreover, public media often yield spillover effects—positive externalities—that encourage better media from commercial and nonprofit sectors as well.

Yet, while salvaging existing public media institutions is important, it’s insufficient for long-term democratic health. A more ambitious project would aim to establish Public Media Centers (PMCs) in every community across the country. These multi-media hubs should be federally guaranteed but locally owned and controlled, perhaps housed in already-existing public spaces such as post offices, libraries, public access media outlets, universities, and public broadcasting stations.

A PMC model can be broken down into six discrete layers that must be democratized: the funding layer determines how this system can be financially sustained; the governance layer ensures resource allocations and other key decisions are made collectively; the ascertainment layer accesses a community’s critical information needs; the infrastructure layer focuses on the material needs (including universal broadband service) for guaranteeing reliable access to information; the technological layer privileges public media in search and news feeds; and the engagement layer empowers local communities in making their own news and telling their own stories. These layers can be seen as part of a broader “public media stack.”

A holistic critique of the problems afflicting our information and communication infrastructures—structural pathologies mostly predating the Trump administration—should inform systemic solutions. Many such problems, including media/tech oligarchy, exclusion from high-quality news and information, and the local journalism crisis all stem directly from or are exacerbated by market failures and run-amok capitalist logics. Only a public media system, if properly designed, can avoid such predictable hazards.

Today, with the ongoing collapse of local journalism and the abject failures of commercial media, the need for a public system is glaringly acute. Any plan for revitalizing American democracy requires an information ecosystem with a fully funded and independent public media system at its core. Reconstructing this system is not a silver bullet for all that ails American society, and other structural reforms remain vitally important. But supporting such essential infrastructure is non-negotiable for any democracy worthy of the name.

I feel that we need a kind of Churchillian response here. We will fight them on the beaches. We will never surrender, and we don’t have institutions right now where we have leaders who are going to say, “Hell no, we are never going to surrender. I don’t care what it costs.” And I feel like we have so few inside of our powerful institutions right now who are not willing to draw that line.

Katy Glenn Bass:

Welcome to Lawyering Without Law, a new podcast by the Knight First Amendment Institute where we explore the unique, and important role that lawyers, and the legal profession play in defending democracy, or facilitating a country’s slide into authoritarianism. I’m Katy Glenn Bass, the research director at the Night First Amendment Institute, and you just heard a clip of a conversation with our second guest, Lawrence Lessig, who we’ll introduce in just a minute. Joining me, and co-hosting this podcast is Professor Madhav Khosla the BR Ambedka Professor of Indian Constitutional Law, and Professor of Political Science at Columbia University. He is also the Knight Institute senior visiting research scholar. Hi, Madhav.

Madhav Khosla:

Hi, Katy. Great to be here.

Katy Glenn Bass:

And with us today, we are very lucky to have Lawrence Lessig, the Roy L. Furman professor of law, and leadership at Harvard Law School to talk about the role of the legal profession in a democratic society. Professor Lessig’s long, and varied career has included work on constitutional law, the internet, and intellectual property, and institutional corruption. He’s published over a dozen books, and for nearly two decades, he has been a leading voice in movements to reform democratic systems, and strengthen institutional accountability. Larry, welcome to Lawyering Without Law. Thank you so much for joining us.

Lawrence Lessig:

Thank you for having me.

Katy Glenn Bass:

It’s a real pleasure. So, I’m going to start us off. You have written extensively about institutional corruption, including in your book, America Compromise, which was published in 2018. So, can you tell us what you mean by institutional corruption, and why it’s such a problem today?

Lawrence Lessig:

Well, I think we have a default understanding of corruption, which is basically illegal corruption, the kind of corruption of taking a bribe, or buying off some government official. And I think that too often we think that that’s the only relevant kind of corruption. And my view is that’s actually the least significant kind of corruption. A much more significant kind of corruption is when we have influences, and in my view, it’s primarily economic influences that guide an institution away from what its underlying purpose is to be. And I think the one we all see is Congress. If you have a Congress filled with people who have spent 30 to 70% of their time raising money so that they can become members of Congress, they’re obviously not raising money from the average American. They’re raising money from the tiniest fraction of the 1%. And so they become extremely sensitive, and responsive to that tiny fraction of the 1%.

Now, our democracy, or our republic was meant to be a representative democracy where members of Congress were to be quote “dependent on the people alone,” as Madison put it. And by the people Madison said in Federalist 57, he meant quote, “not the rich, or than the poor.” But what we’ve allowed to emerge is a dependence on the rich more than the poor because you don’t get to Congress unless you first please the rich. So, in that sense, it’s a institutional corruption. Even if no member of Congress took bribes, still you should say that the institution is corrupt because the dependencies that have developed inside that institution are deeply corrupting from the representative ideals of Congress.

Madhav Khosla:

Professor Lessig, one of the interesting things about the professions is that they each have their sort of internal normativity, and their internal purpose. And so what it means to be a good lawyer is different from what it means to be a good teacher, and a good doctor. And given your focus on institutional corruption as something better understood as structural misalignment, let’s say, rather than illegality, do you think the legal profession meets that definition, or comes close to it at least given how concentrated access to good lawyering actually is?

Lawrence Lessig:

Yeah, that’s a great point. I mean, I do think that it’s first important on the good side to recognize that the legal profession has been pretty aggressive in disciplining its members across its history when they violate some pretty fundamental norms. So, for example, you’ll see all the time that President Trump will say things that his lawyers will not say in court because his lawyers are constrained, they could lose their license in a way that obviously Donald Trump has no license to lose. But I do think that increasingly we’re seeing in the context of especially commercial practice in America, this misalignment. I actually had a personal experience with this when I just came to Stanford, I was asked by a friend to become of counsel to a very prominent law firm. I said, “Your clients all hate me.” And he said, “It doesn’t matter. We’ll take care of that. Don’t worry.” So, I went to work for them, and the very first week one of the biggest clients said, “It’s either Lessig, or us, you need to decide.”

And so they brought me in, and I said, “I told you your clients hate me. This was the point of why I didn’t think this would work.” And they said, “Well, the client is okay if you publish your articles in the Stanford Law Review, or in the University of Chicago law. They just don’t want you to publish in Time Magazine, or Newsweek", back in the day before there was any real internet stuff. And I said, “I can’t make that promise to you. I’m going to publish wherever I’m going to publish.” And then he said, and this is the chilling part, he said, “Well, you’re violating your ethical obligation.” And I said, “What do you mean?” And he said, “Well, you have an ethical obligation to advance the business interests of our clients.” And I said, “No, actually my ethical obligation is to me believing I’m saying the truth, and I have no ethical obligation to your business interest. And so I don’t think I can work here anymore.”

And I left. But the point is, I’m sure there are many who believe that in fact they have an ethical obligation to the business interest of the clients, even if the business interest has nothing to do with the interest of the law. So, that’s a great example of this sort of structural misalignment that I think does exist in the law.

Madhav Khosla:

And it’s funny, your example recalls a conversation Katy, and I were having earlier where we spoke about how lawyers are conceived as being responsible to their clients, but they’re also thought to be officers of the court, and that element has sort of dropped out in some ways in the modern legal profession. But just on Trump’s second term, given the ways in which we find how the legal profession has responded, how have you viewed these events, and do you think that they sort of fit in actually to this broader institutional corruption thesis, and the example that you brought out, and some of the political economy story that you’ve been telling, or do they actually show perhaps some kind of fidelity to the rule of law in some deeper sense?

Lawrence Lessig:

Well, I do think that there’s a really dramatic struggle that’s going on inside of the Justice Department right now where the conception that the president, and his closest advisors have about the role of the Justice Department fundamentally conflicts from the self-conception of most traditional lawyers inside of the Justice Department. Now, the president, and his senior legal advisors are animated by a completely idiotic, historically baseless conception of the executive power of the president, the sort of so called extreme unitary executive power. And under the extreme unitary executive power, the job of everybody in the administration from cabinet officials to lawyers in the Justice Department to judges, immigration court judges, is to be advancing the policy of the president. Whereas the conception of lawyers, especially inside the Justice Department is our job is to advance the rule of law. That’s our job. We work for the Constitution.

We don’t work for the president. Now I’m quite struck by the number of people who’ve said, “Hell no.” So, we can criticize lawyers, and I criticize lawyers a lot, but if you take all these different industries like lawyers, and the media, and tech companies, and you ask in each of these industries, what’s the highest number of people who’ve said hell no? I think we’ve got a lot of lawyers who’ve said hell no. And a lot to have said, “I have a mortgage, and my kid’s going to college so I’m just going to get along, and this guy will be gone in four years. We’ll see what happens.” So, yeah, I think it’s a serious problem. And what I’m optimistic about is that I think more, and more ordinary people kind of see it as a problem.

I’ve been on this corruption gig now for 19 years, and at the very beginning it was very hard to get people animated, or even get them to recognize what the issue was. But I think today nobody can look at this system, and say that the system is immune from that corruption. And of course, the kind of corruption we se today is a million times worse. It’s like Mike in Hemingway’s Son Also Rises, how did you go bankrupt slowly at first, and then all it was, right? And that’s, I think, where we are.

Katy Glenn Bass:

I do admire all of the lawyers who have said, “Hell no.” And right at the beginning of the second term, he issues these executive orders that are targeting a number of big law firms, and a fair number of them, six, or seven, fight those orders in court so far successfully, but then a fair number of them immediately cut deals, including some who were not directly targeted but were trying to avoid being targeted by an EO. So, how do you read that through your institutional corruption lens?

Lawrence Lessig:

Yeah, I think that’s a complete betrayal, totally a complete betrayal. I think it’s because these people have begun to conceive of their job as the job of just maximizing the wealth of their firm. I’m sure you’re familiar with that story that Nick Bostrom pushes about how AI will run away, and he sort of imagines giving a super intelligent AI the job of maximizing paperclip production, and soon we have a whole universe producing paperclips. But corporations were the first paperclip maximizers, and Milton Friedman convinced corporations in the 1970s they had one job. Their job was to maximize paperclip production, maximize shareholder wealth. And that idea even at the time among corporations was deeply contested because many corporations thought, “No, no. Our job is to make sure that our community is healthy. It’s to help support growth in the area.” It’s like a lot of things we’re trying to do.

We’re not just trying to maximize wealth, but that became the kind of cool kids framing for what corporations were supposed to do, and that’s what they became exclusively focused on. And that’s what we’re seeing in corporations today. You say you’re a news organization. No, no, you’re a corporation. Your job is to maximize wealth. So, if you’ve got this tyrant who comes along, and says, “Sell your news department to me, and bend over backwards,” if that increases the shareholder value, that’s your job. You have no choice to do anything else except that. We increasingly see it in other organizations that also have kind of money at their core. So, we certainly see it in the context of these law firms who I think just thought this is simple. We can make this problem go away, and we’ll still have our bonuses at the end of the year.

I think you see that in the news organizations, you see that in many universities. The president comes in basically says, “I’m going to take your academic freedom away.” And they’re like, “Okay, but we’ll at least get the $500 million you are supposed to contractually be giving us.” So, they were willing to settle because they were just focused on the numbers, and I feel like we need a kind of Churchillian response here. We will fight them in the beaches, we will never surrender. And we don’t have institutions right now where we have leaders who are going to say, “Hell no, we are never going to surrender. I don’t care what it costs.” And I feel like we have so few inside of our powerful institutions right now who are not willing to draw that line. I think Harvard kind of tripped on it. I think that the administration was so clumsy, and the demands that they were making to Harvard that Harvard had no choice but to say, “Hell no, we’re not going to accept that.”

And then the administration said, “You caused this problem because you shouldn’t have taken that letter seriously. It wasn’t a serious letter.” And it’s like, “What are we supposed to do now?” But the point is that once they saw the whole world rallied to them, and say, “Wow, you guys are standing up to the president.” It’s like they didn’t realize how much latent energy there was out there in the public to find people who would just do the right thing for God’s sake, stop calculating, just do the right thing. And I think that’s our vulnerability as a society right now. Everybody thinks they’re supposed to do the rational thing. Well, that’s the paperclip maximizer. That’s the end of humanity. And if you don’t have people who are willing to say, “I get this is how I maximize my wealth, but this is how I am human. This is what I do to be human,” we’re lost. We need those people, and we don’t have them right now.

Katy Glenn Bass:

And thinking, going back to this, the 1970s marker that you set out for when the corporate mindset starts to shift, when do you think you see that in the legal profession?

Lawrence Lessig:

I certainly think it’s the corporatization of law firms. I think it’s when the law firms become big, and they essentially are no longer true partnerships where the partners feel this moral, and ethical obligation to each other. And you look at the kind of firms that resisted it. They’re the more traditional kinds of firms. They’re filled with people who think of themselves as warriors. They think of themselves as people standing up for rights, and good. Now, obviously they also have clients who are doing terrible stuff in the world. I’m not saying they’re pure, and then not saying they’re only doing good in the world, but that was their self-conception.

But then you have these firms that are then run by managing partners, and the managing partners are not even really lawyers anymore. They’re just kind of managers. And then managing partners answering one question, how much am I going to give my partners as bonuses at the end of the year, and this strategy’s going to cut that by 20%? I can’t do that. It’s against my job. Well, once that’s what managing partners think, that’s what the law firms think that’s all over for the law firms.

Madhav Khosla:

I guess some of that is the corporatization, and that political economy. Some of it is also actually the spread of law in some sense as law is now trying to serve so many more functions. And so is there some potential story of the idea that lawyers are now existent in this orbit? And that political economy, you mentioned universities, and I was just thinking how it interacts with another profession, universities, and law schools where if you look how much do kids pay to go to Harvard Law School, or to Columbia Law School is a huge amount of money. But I think the answer is because the customer is not the student, it’s the law firm.

And so everybody’s actually locked into that cycle. And the thing that it makes me wonder is that we teach professional responsibility doctrine entirely as being about the individual, but we actually don’t think at all, or teach in any way how to think about this sort of structure. And is it an opportunity to actually redesign how we think maybe about legal ethics, or about just in fact what the profession is as a structural idea rather than just what you need to do [inaudible 00:15:11] in a particular case?

Lawrence Lessig:

Well, I think that’s a great point, and I hadn’t thought about it before, but it connects to the way I would think about Congress as an institution. So, the focus on corruption in Congress was always a focus on whether you had corrupt members of Congress, which of course is an interesting question, but it’s not the most important question. It’s whether the institution itself is capable of acting because of the influences on the institution. And I think that’s a great way to think about the problem of the law as well. Now one big difference though is I’m actually pretty sure that within the next 20 years, we’re going to see a radical change in the institutional structure of law, and for the good. I mean, I’m a believer that as much as I’m terrified about our inability to manage AI risk, and I think AI presents enormous risk.

I mean, I’m not quite a P-doomer of one, but I do think there’s a real P-doom here to worry about. I’m completely acknowledging there’s a totally huge downside. Let’s recognize some of the upside. And one of the most important upsides could be to radically lower the cost of law. If we could spread AI inside of the law, and inside of the administration of justice, and inside of the administration of government, we could just make all of those systems work much better. And what working much better will be is that the rent seeking of these big firms will get blown up, and we’ll see a radical reorganization of what law firms look like. Now, I think there’ll always be lawyers. I’m pretty confident there’ll be many fewer lawyers as the capacity of each particular lawyer goes up because of this technology. So, I’m sure that we’re going to see a radical restructuring of law before we see a radical restructuring of Congress, or maybe not. Maybe Congress is going to blow up next year because obviously there’s zero support for the institution right now.

Katy Glenn Bass:

I’m going to shift gears a little bit here to talk about self-regulation, which you actually spoke about earlier in terms of the ways it is functioning as a real check on lawyers, and their ability to carry out what the Trump administration wants. So, of course, one of the defining characteristics of the professions is that they are largely self-regulating as is the bar, which sets its own ethics rules, and disciplines its members. But you have argued that institutions rarely reform dependencies that they’ve normalized. So, do you see bar self-regulation as something that is structurally capable of functioning as a real check in some of these cases?

Lawrence Lessig:

I don’t think we can speak in general about that, but I’m happy to concede that in general, things are getting more difficult. I do think, again, on the optimistic side, I was very close to Francis Haugen, representing Francis Hagen when she came out as the Facebook whistleblower. And so I spent a lot of time reading the Facebook files. And one of the most striking things to me about reading the files was Facebook’s filled with a lot of super talented, good, sold engineers, and an engineering is a profession. And so here are these good, sold engineers who are going to Mark Zuckerberg, and saying to Zuckerberg, “Look, this technology is killing young girls. We need to do something about this technology", or “This technology is creating extremism, and we could tweak it here, and do this there to make it so that it didn’t do those things.” And Zuckerberg’s singular question, I mean, he was the first paperclip maximizer is “What does it do to engagement?”

And if it reduced engagement, then they weren’t allowed to adopt the reforms. And if it increased engagement, or kept it at the same, they were allowed to adopt the reforms. And so when you think about those engineers, they were not constrained by their profession because their profession doesn’t have any normative constraints on them. You can say you’re supposed to be acting in a certain way, but there’s not an institution that takes away your engineering license, right? There’s nothing there to police them like that. And so I had many conversations with engineers who are kind of desiring to create the equivalent of like the ABA for engineers, and set a bunch of ethical codes for engineers so that engineers are able to point to the sign that used to be on Google’s law, which is do no evil. We’re not allowed to be an engineer, and do harm like this.

And that would be an enormous advance. I spent many years thinking about the way architecture, or code is law. Well, if that’s true, the engineers ought to have the same kind of professional constraints on them that lawyers have on them to think not just about what their client wants, but what they understand the values of engineering to be. So, I think that that’s an optimistic way to think about how that profession could be regulated. But I think in the American context now that dominance of these big firms is distorting the capacity of this normative rule to have a effective constraint on how lawyers in general are going to function, and I’m not sure what’s going to change that in the short term.

Madhav Khosla:

Larry, on your big picture point, and zooming out maybe from this example, just on the relationship between lawyers, and democracy, let’s say. One of the things that I’ve often thought about with your work on Congress is that you correctly focus on the number of people in Congress who spend a huge number of their time raising money. The other interesting thing about Congress is that lawyers make up a disproportionate share of it. And I was wondering whether, given that a lot of your work is on how the normativity, and the logic, and the principles of one domain sort of seep into another, do you think legal training from adversarial thinking, formal compliance, client loyalty makes lawyer legislators actually perhaps worse at recognizing the kind of corruption you described? Because lawyers also assimilate quite well into new structures, or path. There’s this fabulous line in Tocqueville’s Democracy in America that in a society where lawyers don’t get assimilated into structures of power, they will be active agents of revolution.

Lawrence Lessig:

Yeah. I first of all think that we have wildly too many lawyers in government. I also think that one thing we know about intelligence is that the smarter you are, the more you can convince yourself of whatever you want to convince yourself of. When they do studies of climate deniers, they don’t find that it’s uneducated people are the strongest climate deniers. It’s actually the educated people who are the strongest because they’re very good at rationalizing to any conclusion. And so lawyers are kind of trained at that game of rationalization. And so yes, if you have a lot of people whose profession is to practice this kind of rationalization, then you’re going to have that distortion. I mean, this is one particular, I think, corruption of American legal practice relative, for example, British legal practice. One great thing about the British system is that the advocates, the advocates who show up in court will argue on both sides of criminal practice.

They will be prosecutor one day, and defense the other day. And I think that discipline helps them understand the contours of a really just system. Whereas in the American system, you go down one path, you can never cross over. You do criminal defense work, you’re never going to be a prosecutor. You’re a prosecutor, you’ll never be touched by anybody in the criminal defense industry. So, that’s a really bad thing, kind of amplifies this bias. But absolutely. I mean, if you change the way you fund elections, which I think is ultimately what we have to have, you could increase radically the number of women, and the number of middle class, and diversity economically of people inside of Congress, because right now the natural successful person running for Congress is a white male lawyer that’s the perfect person to run for Congress. Why? Because they have what we used to call Rolodexes.

The biggest Rolodex is they have all of these clients who are going to obviously be interested in them being in Congress, and they can call all of them, and they can get all the money they need to be able to run for Congress. But if you’re like a working mother, or if you’re somebody who’s been working in a steel mill to the extent there are steel mills anymore in America, you can’t compete with that. You have no opportunity to compete with that. But where we’ve seen the system move to public funding systems, my favorite example is Seattle, which has vouchers. Basically all voters get vouchers, and they give the vouchers to candidates, and candidates fund their campaigns based on the vouchers they get from voters.

We see a significant increase in women, and extreme increase in the economic diversity of candidates, and extreme increase in the cultural diversity of these candidates. So, you begin to have candidates who reflect the society as a whole, and I think that’s the critical thing that we have to do. And if we did that, there would be fewer lawyers, and that would be, I think, a good thing.

Madhav Khosla:

Larry, one of the things about you is that you inhabit, in some sense, two professional lives, right? You’re a lawyer, and you’re a teacher, and you are part of a group of colleagues who co-wrote a letter in March of 2025 about the role of legal education, and lawyers with respect to democracy, and the rule of law. And can you just tell us just a little bit about that, and the way your students felt about it?

Lawrence Lessig:

Yeah. I mean, I think we can’t minimize the extent to which this administration is abnormal. And I think the question that many of us felt is, do we remark its abnormality as a way to signal that this is just not the way law is supposed to function? And I think none of us, even in March of 2025, had any clue about just how extreme, and abnormal it would become. And so I think we tried to articulate not an anti-Trump letter, but a letter that was emphasizing basic commitments to principles of rule of law that the administration was deviating from. It was a difficult experience with many of the students. I mean, I wouldn’t say a majority, it’s the Harvard Law School, but there were a significant number of conservative law students who were angry about that. And my colleague, Adrian Vermeule, in his Twitter face at least, was quite angry that the faculty had signed that letter.

But I pushed back both on Adrian, and also on some of the students, and I said, “Just tell me exactly how our conception of the rule of law is wrong. What have we said that’s just not true?” Because if you’re saying we shouldn’t say it because it’s against the existing administration, I just don’t buy that. And so when you think that Joe Biden was slapped on by the Supreme Court under the Major Questions Doctrine because his policies deviated from the traditional position that the law had taken, and then you compare that to this administration, and what the court has done in this administration, it’s night, and day, and people say the best parallel for Donald Trump is Franklin Roosevelt. And I think that’s true from the perspective of how significantly has he changed the law. But what’s different about it is that Franklin Roosevelt had Congress behind him, and then he had a midterm election where he overwhelmingly swept the country, and he had an even greater support after that.

And so there he had the people behind him as well. Here, we don’t have any laws of Congress at the core of the fights that are going on in the Supreme Court, and we certainly, assuming we have an election, not going to have a midterm election that rallies behind Donald Trump. So, he has grabbed more power than any president in the history of the nation, even if he might have not have as much additional power as Franklin Roosevelt, but Franklin Roosevelt didn’t just grab it, he earned it, and he did it along with the Congress, which is not happening now. And that’s why I think it’s the job of the court to be willing to lose in the face of the president. Those of us who study this look back to the court in 1935, it struck down critical laws that FDR had gotten past.

And most of the time people look at that, and they say those old conservative justices, they were just like imposing their own views. But I think what they did was really important, and very valuable because what they did was draw a line in the sand, and say to the public, “You tell us which side you’re on because we’re telling you what we think our tradition means, and whether we’re right, or not, what this president is doing is inconsistent with the tradition.” And that triggered a debate in the public, and that debate in the public was democratically edifying. Right now we’ve got the court basically ducking, and weaving wherever they can to avoid any real conflict, and we don’t yet have them sort of triggering a public reaction that says, are we with Trump in his new conception of us authoritarian presidency, or are we going to stand for what America has traditionally been? And so I think there’s a lot of confusion that’s left because of that lack of clarity, which I think we are all suffering for.

Katy Glenn Bass:

And still thinking about sort of shaping young lawyers, and their conceptions of themselves as lawyers in relationship to the rule of law, and to upholding democratic values. So, to what extent do you think legal education is currently doing a good job of this? To what extent can legal education even accomplish this goal? Is there a way to reshape legal education that might improve the way lawyers emerge, and their understanding of their role in upholding a Democratic system along with their clients’ needs?

Lawrence Lessig:

Yeah. So, I feel like I’m probably the least informed person to answer that question. I can tell you what happens in my class, and you’re not going to be surprised that I think what happens in my class is good because I feel like in my class, what I try to do is to get people to reflect all the time on the commitment to a principle over the expediency of the moment, and that decision that they need to make. And that’s as much about their own personal choices. So, like when I counsel law students, I will tell them the really important thing you need to do is to make sure you’re never spending more than 60% of your salary. Why? Because you always want the freedom to choose to do something differently. And if you spend 90% of your salary, you’re locked in, you can’t change.

You’ve got like a mortgage, you’ve got kids in school, and that the firm does something you don’t like, you’re stuck, but you need to maintain that freedom because you’ve always got to be making the decision. Am I doing what I believe in? Am I doing what I can have confidence is right? And so it’d be great to talk to law students who’ve gone through it, and see how often do they feel like they’re being called on to make that judgment, and to decide on which side of that division they’re going to stand. Now, the other great advantage I have is I teach first years, and first years are the most idealistic. These are people that are going to change the world by doing right, and like arguing for what’s right. And so I just love the opportunity to teach people with those ideals, because it’s easy to just push a button with them, and get them to reflect, and to see this is why I came.

This is why I wanted to be a lawyer. I wanted advance causes that I think are valuable, but I’m sure that by the time they graduate, the system has worn that out of many of them, and they’ve accepted that they’re going to go work for a law firm, and they’re not going to be doing the kind of practice that necessarily advances their conception of justice. And the truth is between those two extremes, but I would love to understand how our students experience this more.

Madhav Khosla:

As we move towards closing this, and circling back a little bit, one of the things that you said that was very powerful that’s been playing at the back of my mind is the capacity for self-rationalization, and how it’s not fortuitous that lawyers would be so good at it. Citizens United has been so central to your work because it’s both exemplified, and made worse a problem that you have had your finger on for years. If you think about something like that, and you think about it as a dramatic political crisis, it’s also true that it was built by lawyers. It was decided by a court of lawyers justified through a First Amendment framework that lawyers developed. If we think about this, there is this one moment where let’s just say the corruption of the legal profession exists because of a certain political economy, that law firms have become basically corporations.

And so the fidelity is not actually to something law related, but the fidelity is just to maximizing profit. But there’s another where it may not actually even be necessarily maximization of wealth, or the corporate mindset in that way, but actually just instrumentalism, or careerism, or just the idea that actually the only thing that matters is clients, and winning, which is a somewhat distinct kind of problem to just money. And it seems very pervasive because part of what is interesting throughout in the professions is that professions are almost definitionally non-instrumental. There’s something about what it means to be a good doctor, which can’t be captured not only by money, but not by anything instrumental in some sense.

Lawrence Lessig:

Yeah, I think that’s a good point. I think that it’s important to recognize the way lawyers can be exploited by interests that are trying to advance their own interests, not for principled reasons, but just to advance their interest. I often think about Robert Borick who began much of his work was around constitutional theory, but he became really famous for his crazy views on antitrust. When he wrote his first piece on antitrust, it was so crazy that nobody at the Yale Law School took it seriously. It was completely removed from reality. But Borick was just sort of spewing out ideas, constitutional ideas, antitrust. He was just an academic. And all of a sudden a certain set of them became really valuable to economic actors, and they got amplified. And when they got amplified, they became quite significant in the world. I also think that’s kind of true in the money, and politics area.

There are people who absolutely for good faith reasons are advancing a very extremist view about the libertarian conception of campaigned finance, where you’re not allowed to have any regulations at all. That produced Citizens United. But from my perspective, the real frustration about Citizens United is the extremely bad lawyering that has allowed people to believe that Citizens United created Super PACS because Citizens United did not create Super PACS. Super PACS were created by a lower federal court decision, Speech Now versus FEC, three months after Citizens United. And that decision is based on a very basic obvious, once you see it, logical mistake. But once Speech Now made that decision, three other circuit courts followed it within a couple years, and so it became settled law. Now we’re in the middle of a fight about this right now. We’re in the First Circuit. We got an initiative passed in Maine by 75% of the vote banning Super PACS in the First Circuit.

We’re going to have an argument as soon as it gets scheduled about whether that position is actually mandated by the Constitution. The Boston Globe just had an editorial yesterday where they said the court’s got to take this up right away because it’s really critical, and they’ve identified the mistake, and there’s a chance the Supreme Court will do the right thing. And in the Supreme Court, I feel like we will be giving the Supreme Court a gift because we will be saying to the Supreme Court, “You don’t have to reverse Citizens United. We’re not asking you to reverse Citizens United. We’re asking you to apply the reasoning of Citizens United to a set of facts that you’ve never considered because you didn’t review that decision to give us Super PACS.”

And I think the Supreme Court’s going to be able to write an opinion that says, “Hey, Bernie Sanders, you’re wrong. Citizens United is perfectly fine. We totally endorse Citizens United,” but there’s no reason why Citizens United means that you have the right to contribute unlimited amounts of money. And so Super PACS can be limited even if the expenditures cannot be limited. And that will be a huge victory, and it will also be a huge embarrassment to the legal profession because we’ve lived with this deeply corrupting system of Super PAC spending for the last 16 years. And there’s no reason that we should have taken... It’s going to take at least 18 years to get to the place where the Supreme Court fixes it. And the idea that we’ve lived for 18 years with what is in essence a logical, simple, obvious mistake is an embarrassment. A legal system that kind of makes that kind of mistake is not a system you should be proud of.

But I think that I’m pretty confident we’re going to get this corrected, and when that happens, we’ll have a reason to celebrate the balance that can come. But I think that the reason we’ve not seen that is that all the money in the world is on the side of the libertarian position, and it’s really been hard to stand up something in balancing that position that has a fighting chance of prevailing.

Katy Glenn Bass:

Okay, that’s fascinating. What is the name of this case so that we can start tracking it?

Lawrence Lessig:

So, the case is DTA Dinnertable Action versus Schneider, who’s the head of the ethics commission in Maine. And we recruited Neil Katial to be our lawyer. So, Neil, I’ve known forever since he was a student, and I was trying to get him, and he was avoiding my calls, and finally I got him on the phone, and he said, “Lessig, look, I don’t take loser cases. It’s a total loser. You’re never going to win this case. I’m not going to waste my time on it.” I said, Neil, just have lunch with me, just one lunch. And so he took me to the worst restaurant in Washington, and we’re having lunch. And midway through the lunch, he said, “Oh, my God, you’re right. You’re right. I can win this case. I can totally win this case.” And so I think that when Neil, the tariff slayer gets up there, and tries to slay Super PACs, we’re just going to call him the slayer. He’s like the tariff slayer, the super PAC slayer. He’s just the most impressive lawyer to persuade conservatives to do the right thing that we have. And I think he’s going to prevail.

Katy Glenn Bass:

That’s really exciting. It’s like we have a legal cliffhanger to end this episode on. So, we will pay attention to that. So, you’re awaiting scheduling in the first circuit for this case.

Lawrence Lessig:

That’s right.

Katy Glenn Bass:

Okay. We will keep up with it. Larry, this has been really great, and really thought-provoking. Thank you for taking the time to come on the podcast, and sharing your views, and your long years of scholarship, and insight with us.

Lawrence Lessig:

Thank you for having me.

Madhav Khosla:

It’s been a privilege, Larry, really. Thanks so much for being on this.

Lawrence Lessig:

Thank you, Madhav.

Katy Glenn Bass:

Lawyering Without Law is a production of the Night First Amendment Institute at Columbia University, and is hosted by Madhav Khosla, and me, Katy Glenn Bass. This episode was produced, and engineered by Dustin Foote, fact checking by Connor Menzies, and Sophia Rojas. Candace White is our executive producer. Our music comes from Invato Elements. The art for our show was designed by Jay Volmar. Thanks to Lawrence Lessig who joined us for this episode. Lawyering Without Law is available on Apple, Spotify, and wherever you listen to podcasts. Please subscribe, share, and leave a review. We’d love to know what you think. To learn more about the Knight Institute, visit our website, knightcolumbia.org. That’s Knight with a K, and follow us on social media. We’ll see you next time for a conversation on the shifting identity of the legal profession with Professor Deborah Pearlstein of Princeton University. Bye for now.

1. The Business Model Problem

A healthy information ecosystem requires actors who investigate and report on the activities of government officials and others—in a word, journalists. Beginning with the launch of Craigslist in 1995 and accelerating with the rise of advertising on Google and social media platforms, classified ads, which traditionally accounted for thirty percent of the revenue for local newspapers, dried up. Subscriptions also declined, as erstwhile readers turned to the internet for free alternative sources of (often unreliable) information. These and other factors hit local journalism the hardest, creating local “news deserts” that have been widening for two decades.

One partial response is public funding. And indeed, for a long time, that was an important piece of the puzzle, especially in rural areas, where funding from the Corporation for Public Broadcasting (CPB) accounted for a critical portion of local television and radio stations’ revenues—as high as 97 percent in one rural Alaskan community. However, Congress eliminated CPB funding in 2025, and with nothing in the till to distribute, earlier this year CPB’s board voted to dissolve itself.

If and when Congress and the administration change hands, it will be tempting to resurrect CPB or re-create something like it. I would not oppose such a move, but it carries risks. In stable constitutional democracies, publicly funded news services such as the British Broadcasting Corporation (BBC) and the Canadian Broadcasting Corporation (CBC) make a healthy contribution to public discourse, providing honest and nonpartisan news reporting. However, in the hands of an autocrat, government-funded news organizations typically serve as sources of propaganda.

Seen in this light, the elimination of CPB funding, while harmful, was not as harmful as one of the alternatives: continued funding but on condition that the recipients of government largesse toe the party line. Such moves can be and have been used even while the authoritarian regime purports to respect the independence of journalists. Examples from other countries readily come to mind, but so do domestic ones. When the Defense Department began exerting greater control over Stars and Stripes earlier this year, it claimed that the news organization would continue to “operate with editorial independence,” while it underwent a “modernization” under which it would no longer focus on “woke distractions.”

To be sure, it is probably easier for the government to exert control over a house organ like Stars and Stripes than over independent local television and radio stations, but only slightly. President Trump filed frivolous lawsuits against ABC News and CBS News but nonetheless obtained multi-million-dollar settlements because they feared that his administration would otherwise withhold regulatory approvals vital to their business. He or a similarly autocratic successor would have little difficulty finding and using leverage over much smaller news organizations.

The upshot is not necessarily that a future, less autocratic, government should refrain from funding journalism. Rather, the point is that any such efforts should be undertaken cautiously, with eyes open to the risks. Lawmakers should consider mechanisms to mitigate those risks, such as independent funding streams not subject to presidential control. Finding such mechanisms will be difficult if, as widely expected, the Supreme Court invalidates nearly all independent federal agencies in Trump v. Slaughter. Reliance on state rather than federal funding would be one possibility.

2. A New Kind of Fraud

The decline of local journalism is not the only adverse impact of the internet on our information ecosystem. It has also led to the spread of misinformation and disinformation. Any successful effort to combat that phenomenon would necessarily be multi-pronged. Here I would suggest one prong: expansion of liability for fraud.

In most U.S. jurisdictions, civil or criminal liability for fraud is established by showing a deliberately false representation that is intended to and does induce reliance by the listener, causing damage to the latter. In some statutes, there is also a requirement that the fraudster obtain money or some other thing of value from the person defrauded. Yet despite the fact that we take part in an “attention-based economy,” purveying mis- or disinformation in exchange for monetizable attention has not been regarded as a species of fraud.

A 2022 student Note proposed that the Federal Trade Commission (FTC) could regulate “fake news” as a species of fraud that falls within its existing statutory mandate. But whether spreading mis/disinformation would be regulated by the FTC, made the basis for civil liability via common law actions, or defined as criminal or tortious by a new state or federal criminal or civil statute, liability would be subject to First Amendment limits. “Simply labeling an action one for ‘fraud,’” the Supreme Court stated in a 2003 case, does not suffice to take it outside the protection of the First Amendment. That’s fair enough. However, deliberate mis/disinformation that aims at capturing the readily monetized commodity of attention bears sufficient similarity to conventional fraud as to fall outside First Amendment protection.

I do not have the space here to address any of the details that would be necessary to implement a conception of fraud capacious enough to cover mis/disinformation. For example, to hold platforms liable, some amendment to Section 230 of the Communications Decency Act might be required.

Nor can I ensure that liability for mis/disinformation would not be weaponized by the likes of Trump or his political allies, who frequently claim that truthful reportage that puts them in a negative light is “fake news.” But that risk would be worth taking if we concluded that the autocrats already have more potent tools at their disposal.

Defamation liability is arguably such a tool. For example, for FBI Director Kash Patel to prevail in his defamation lawsuit against The Atlantic, which ran a story describing him as engaged in “bouts of excessive drinking” and other irresponsible conduct, he would need to show that the reporters and/or writers for The Atlantic acted with reckless disregard for the truth. That is the high bar set by New York Times v. Sullivan, but it is still lower than the bar for fraud: intentional falsehoods. However, there are other elements to a successful defamation claim, not least that the statement at issue be damaging to the plaintiff’s reputation.

By contrast, liability for fraud for mis/disinformation about a potentially limitless number of subjects could open the door to many more claims than defamation law now permits. Accordingly, liability for mis/disinformation as a species of fraud should probably be limited to some discrete and especially dangerous subset of such mis/disinformation, such as AI-generated images, videos, and sounds passed off as real.

* * *

Vincent Blasi has argued that the First Amendment should be interpreted from a “pathological perspective”—that is, to equip it “to do maximum service in those historical periods when . . . governments are most able and most likely to stifle dissent systematically.” I agree. Legal regimes that may be sensible in placid times can prove hazardous in authoritarian periods, such as the current one. However, the pathological perspective should be supplemented by an optimistic perspective that not only guards against autocracy but also fosters conditions for a robust democracy.

This workshop and essay series are part of a broader project on the legal profession and authoritarianism, organized by the Knight Institute in partnership with our Senior Fellow Madhav Khosla, the Dr. B.R. Ambedkar Professor of Indian Constitutional Law at Columbia University. This project seeks to spark public debate and generate new scholarship that will galvanize members of the legal profession in the United States and abroad to reflect on their roles in defending the rule of law and democratic values, and help them be more effective at doing so. It also includes a new biweekly podcast that interrogates the unique and important role that lawyers play in defending democracy, or in facilitating the slide into authoritarianism, co-hosted by Khosla and the Institute’s Research Director Katy Glenn Bass.

The essay series will explore, among other questions, how differing conceptions of the legal profession contribute to lawyers’ role in either resisting authoritarianism or accommodating it, how bar associations maintain—or lose—their independence in inhospitable political environments, and how democratic backsliding can occur through the working of ordinary legal rules and practices. We are excited to announce that the following scholars and jurists will participate in the September 2026 workshop, and many of them will contribute essays to the series: 

Lukman Adebisi Abdulrauf, University of Ilorin

Payam Akhavan, University of Toronto

Aslı Bâli, Yale University

John Coates, Harvard University

Scott Cummings, University of California, Los Angeles

Max du Plessis, Doughty Street Chambers, Senior Counsel, South Africa

David Dyzenhaus, University of Toronto

Vicki C. Jackson, Harvard University

Jeff King, University College London

Daniel Markovits, Yale University

Pratap Bhanu Mehta, Princeton University

Deborah Pearlstein, Princeton University

Margaret Satterthwaite, New York University

Kim Lane Scheppele, Princeton University

Michael Sfard, Michael Sfard Law Office, Tel Aviv

Kate Shaw, University of Pennsylvania

Rachel Stern, University of California, Berkeley

The spectacle of universities, law firms, media companies—even as venerable an institution as the Smithsonian—bending to the administration’s speech-suppressive pressure obviously should be of real concern to all who care about democracy in the United States, even if it is not permanent. It may be that, as the Trump administration becomes less popular or overplays its hand, civil society organizations show more willingness to resist the government’s mandates. The free speech crisis we are currently experiencing may to some degree solve itself. But it may not. And regardless, the past fifteen months have illuminated very serious structural vulnerabilities in the U.S. system that lie waiting to be exploited by other, perhaps cannier, government actors. We thus need to understand why the system proved so susceptible if we are to have any hope of making it less susceptible going forward.

The evidence of recent history suggests, perhaps counterintuitively, that the primary locus of blame does not lie with the bodies of substantive law—and in particular, the First Amendment—that are nominally charged with constraining the executive when it comes to the regulation of speech. This is not to say that First Amendment law could not do more to protect the independence of the democratic public sphere than it does at present. As I have detailed at length elsewhere, government speech doctrine has provided the administration a powerful shield against First Amendment challenges to its efforts to remake speech and association in the federal bureaucracy. Meanwhile, the deregulatory impact of First Amendment campaign finance law has clearly limited the administration’s vulnerability to political, as opposed to legal, checks on the president’s power.

But it is also true that, over the past sixty years, and particularly in the wake of the Second Red Scare, courts have interpreted the First Amendment to strongly protect private actors against many of the speech-suppressive tactics (jawboning, retaliation, the conditioning of federal benefits) that the administration has tended to use. And yet, institutions targeted by the Trump administration have chosen not to invoke these legal protections to defend their own expressive freedom and the freedom of those they employ, educate, and host. That pattern, which has repeated itself across institutional contexts, points toward a conclusion that is as important as it is uncomfortable: the vulnerability the administration has exposed is not primarily a failure of doctrine but a failure of institutional incentives.

What this means for those who wish to address the structural weaknesses the past fifteen months have laid bare is that the solutions cannot lie simply in reforms to First Amendment doctrine. The project of free speech reconstruction also needs to pay attention to the institutional, legal, and economic factors that led so many institutions—and more specifically, institutional decision-makers—to believe it rational, desirable even perhaps, to give in rather than to use the tools at their disposal to fight. Put differently, if we have learned anything from the second Trump administration, it is that the health of the American system of free expression depends not only on what restrictions the law imposes on government power but on what private institutional actors are incentivized to do when the government exercise its power, in either legal or extralegal ways, and on whose incentives matter when it comes to institutional decision-making.

One of the more instructive patterns to emerge from the past fifteenth months is the divergence in the willingness to fight between senior institutional decisionmakers and those in the same institution who were more directly involved in its everyday speech hosting and disseminating activities. In law firm after law firm, university after university, media organization after media organization, the people who decided whether the institution should comply or resist were, with notable exceptions, not the people whose professional lives were most immediately organized around the speaking and listening the institution fostered. The attorneys who litigated for disfavored causes, the faculty who pursued heterodox research, the journalists who reported on inconvenient subjects—these were the institutional actors who had the most direct stake in the outcome, and who, in many cases, showed the most willingness to push back. This is not terribly surprising. As decades of scholarship on the problem of proxy censorship suggest, those who host the speech of others tend to weigh the free speech costs of compliance with government censorship less heavily than the speakers or associations they facilitate. This is for an obvious reason: it is not their speech that is on the line.

This structural misalignment between those who make decisions for free speech institutions and those who bear the expressive consequences of those decisions points toward a reform agenda that differs in important respects from the one that has dominated free speech scholarship in recent decades. The central task, from this perspective, is not simply to construct stronger external legal constraints on government overreach (though that obviously remains important) but to restructure the internal governance of the private institutions that comprise the democratic public sphere, and the economic and political markets in which they operate, so that those who make decisions on behalf of those institutions are more meaningfully accountable to those within the institution whose expressive and intellectual work the institution exists to enable. The mechanisms for doing so will vary by institutional context. In some settings, existing institutional structures—for example, the norms of shared governance at public and private universities, and in a different way, law firms—may offer promising points of intervention. In others, reform to the external institutional environment—for example, ownership restrictions that prevent the consolidation of control over media markets—may be more promising. What the mechanisms have in common, however, is that they operate not by limiting what the government can do but by altering how it is that the targets of government power react.

This may seem like an overly modest way of approaching what feels like a large and urgent constitutional crisis. But if the free speech system's vulnerabilities are rooted as much in the incentive structures of private institutions as in the adequacy of constitutional doctrine, then addressing those vulnerabilities will require reforms that work at the level of institutional architecture, and that treat questions of organizational governance as questions of free speech theory. That is a different kind of free speech scholarship and a different kind of free speech advocacy than the field has typically produced. But what the past fifteen month strongly suggest is that without this kind of institutional reform, even the most beautifully designed First Amendment rules will be unable in practice to safeguard the independence of the public sphere from governmental sticks and carrots.

The following can be attributed to Katie Fallow, deputy litigation director at the Knight First Amendment Institute at Columbia University:

“This is the latest attack in the Trump administration’s war on press freedom. Time and again, the administration has shown itself willing to disregard the First Amendment and long-standing limits on the use of government power to go after news outlets that publish embarrassing or critical information about the government. Subpoenas targeting journalists can deter reporting that is vital to our democracy. The Justice Department should explain publicly why these subpoenas were necessary and legally permissible, and Congress and the courts should scrutinize that explanation carefully.”

For more information, contact: Adriana Lamirande, adriana.lamirande@knightcolumbia.org 

Our debates about ‘AI’ grow out of 1990s science fiction. Back then, Vinge (1993) wrote essays and novels urging us to face up to the oncoming “Singularity”: a moment of rapid change that would fundamentally transform the human condition. On that day, AI would rapidly evolve from merely human-level intelligence, what some now call ‘artificial general intelligence’ (AGI), into something super-intelligent with its own interests and goals. Humanity would then either be casually eliminated by out-of-control machines, or humans would become as gods, with super-human servitors at our command.

However excellent the resulting science fiction (especially Vinge (1992)), it is a scandal that this dream of the ’90s is still alive and shaping debate. For complicated social and cultural reasons (Becker, 2025), many of the progenitors and funders of modern generative AI bought heavily into this mythology, and built their business strategies and innovation around it (Hao, 2025). As Singularity thinking has leaked out of containment, it has fueled speculation about further vast social, political, and economic transformations. Will AI supercharge authoritarian mind-control (Harari, 2018) or remake democracy (Gudiño et al., 2024)? Will neoliberalism become a feral, self-aware, and all-devouring “machinic” system (Land, 2011)? Again, there are excellent science fiction treatments of these and other possibilities (Banks, 1987; Reynolds, 2000; Stross, 2005; Chiang, 2010; McAuley, 2010; Valente, 2011; Emrys, 2022), but novelists are (typically) more careful about incorporating complexities, and few set themselves up as prophets, or even prognosticators.

Authors of speculative non-fiction about AGI are less inhibited, offering sweeping visions of how information technology will completely transform society, economy, politics, or all three. They treat AGI less as a technology than as Andreessen (2023) says, “our alchemy, our Philosopher’s Stone,” an alkahest that will dissolve the dross and cruft of human institutions, leaving only pure, undiluted progress. Prognosticators regularly describe the social institutions that tremble on the brink of transformation in ways that are only slightly less stylized, claiming for example that AI might transport us into Condorcet’s utopia of a new Age of Reason, this time, happily without tumbrils of condemned prisoners waiting on their appointments with the guillotine (Hall, 2026). The result is a genre that we believe is more liable to confuse smart people and lead them astray than to usefully guide public action.

All of these aspirations and arguments are, as we said, rooted in myths which are (at least) twenty years older than the technology which now seems to incarnate them, large language models (LLMs). It is because LLMs moved in the space of a few years from being a technical improvement in machine translation (Vaswani et al., 2017) to being proclaimed as the royal road to AGI that these debates really matter. LLMs are remarkably good generative statistical models of human language (including human-written computer code). This allows them to process language in ways that resemble human discourse and to be jury-rigged to create texts that loosely approximate human reasoning. This is a new material reality, a new force in the world, but one whose actual implications are obscured by the mythic garb it is swaddled in.

If we are unimpressed by stories about paperclip maximizers remaking the galaxy, omniscient bureaucracies of terror or wonder, markets that suddenly become self-aware, and the like, it is not because we think they are too weird. Rather, they are not nearly weird enough, and miss how much of the weirdness is already here. The possible futures we face are much messier and more varied than stark visions of omnipotent AGI, just as our immediate past was. They will be shaped by the collision between imperfect and highly complex technologies and imperfect and highly complex human social systems (Matias, 2023; Nelson, forthcoming). It is impossible to predict the consequences, but we can map, study, and think about them as they are happening.

From our perspective, the Singularity began two centuries ago with the Industrial Revolution (Shalizi, 2010), and it has been much messier and more variegated than anyone could have known. The modern social sciences are the offspring of the enormous shocks that they entailed in the past (Tilly, 1984; Nelson, forthcoming). They now need to work together with computer science and other related disciplines (science and technology studies; communications) to map what is best grasped as another stage in the Long Industrial Revolution. AI may turn out to be very important, but in quite different ways than our inherited myths suggest.

We build on ongoing collaborative work (Farrell et al., 2025) with Alison Gopnik and James Evans which argues that it is a category error to think of “large models” as self-motivated agents in the making. Instead, they are better understood as “cultural” (Yiu, Kosoy and Gopnik, 2024) and “social” technologies, resembling libraries and languages on the one hand and markets and bureaucracies on the other. Here we focus on how to study these technologies’ consequences for human society, emphasizing the social rather than the cultural aspects. We particularly emphasize how AI is a social technology, a systematic means of reorganizing social relationships among human beings (Therborn, 1978). Earlier social technologies include not just other information technologies, but institutions of governance such as bureaucracies, markets, and even democracy (Farrell, 2025). We will focus on LLMs over other AI systems. This downplays some important aspects of modern AI (e.g., its use in straightforward scientific problems such as protein folding) but helps highlight connections to other social technologies.

Briefly: LLMs create social relations between their users and the authors of the text in their training corpora. With the right access to the model and the corpus, one can trace the connections from system output back to individual source texts and their authors (Grosse et al., 2023). These social relations are mechanically mediated, giving users the illusion that they are interacting with just the machine and not an assemblage of people. But mediated social relationships and their illusions are a common fact of modern life. The social relations created by LLMs in turn cut across, and interact with, other social relations, including those shaped by other social technologies.

Our goal here is to clear a common space where the social sciences and computer science and engineering can discuss the social consequences of AI. We draw heavily on the ideas of Simon (1996), who saw AI, political science, administration, economics, computer science, and cognitive psychology as so many branches of the “sciences of the artificial,” studying how human beings create "artifacts" that model, and act on, their environment. From this perspective, AI models are another means of “complex information processing” (Newell and Simon, 1956). As Simon emphasizes, such systems encompass both information technologies, as studied and built by computer scientists and engineers, and social information systems such as markets, bureaucracy, and, although Simon himself does not stress this, democracy (Lindblom, 1965). All such systems process information by reducing complex realities into more tractable ‘coarse-grainings’ or abstractions that (hopefully) capture important features of the data. Producing coarse-grainings is not all that large-scale social institutions do, but it is quite important. Economic, administrative, and political coordination simply cannot work at scale if complex social relationships are not compressed into visible, tractable representations.

This then opens a different perspective on the collision between new technologies such as AI and existing social systems. As DeDeo (2017) suggests, we urgently need to discover how the new coarse-grainings of AI interact with the existing abstractions through which humans simplify an inherently complex world to make it tractable. Both AI and older social technologies are, among other things, forms of information processing. We should investigate how the former are variously reinforcing, reshaping, or replacing the latter.

In the rest of this paper, we briefly situate AI in the historic context of the Long Industrial Revolution. Next, we explore the relationship between social technologies and coarse-grainings or abstractions, emphasizing their lossiness and consequences for power relations between different social groups. That allows us both to describe the apparent strengths and limits of actually-existing AI and start applying Simon’s ideas to the intersection between AI and bureaucracy. Bureaucracy is a crucial, ancient social technology which played a central role in Simon’s work. Its relationship to AI is urgently topical: claims about AGI were seemingly one influence on the Trump administration’s sweeping cutbacks to the administrative state. We contrast these ideas with our own, to draw out the many important questions and problems that are elided or ignored by AGI-fueled speculation. Rather than expecting AGI to resolve perennial problems of human social organization, we should treat AI as a new social technology which will alleviate some problems, exacerbate others, and create new ones, just as other social technologies have done in the past. That, in turn, suggests the urgency of cooperation between social and computer scientists to figure out its social consequences, and broader social and political coordination too, of the kind that happened in previous stages of the Long Industrial Revolution.

Coarse-Grainings and the Long Industrial Revolution

We begin from a different viewpoint on the relationship between AI and society than much existing commentary. Scholars of science and technology studies (STS) are often more interested in how scientific and technological systems reflect broader social and political power relations (or develop their own) than in providing detailed social science microfoundations for their arguments. In contrast, the ‘rationalism’ that has dominated internal debates over AI is microfoundations all the way up. It starts from the assumption that the relationship between human beings and AI agents can be understood through the micro-level lens of strategic competition among rational Bayesian agents, and has only recently begun to think systematically about how collective phenomena might emerge as these systems scale (Hammond et al., 2025).

Herbert Simon’s intellectual project differed from both. His work provides microfoundations for an account of larger social institutions, which is explicitly grounded in the ‘bounds’ to individual human rationality (Simon, 1957). Simon suggests that large scale social technologies emerge from the need of limited humans to build collective arrangements that allow them to map and manage a complex world. Simon’s consistent theme is the mismatch between the complex environments that human beings inhabit and remake, and the limited information processing capacity they have to understand it. Here, Simon’s view partly converges with Dewey’s understanding of democracy (Farrell and Han, 2025). Neo-classical economics tends both to smooth away the complexities of the environment, and to assume that individual humans have unlimited computational power to model it and find optimal solutions to their problems. These assumptions are both mathematically convenient and highly unrealistic. From Simon’s perspective, humans must usually satisfice rather than optimize—settling for ‘good enough’ solutions rather than the best possible. Simple mental heuristics can help discover such solutions. So too can social institutions that channel the production and gathering of knowledge across many individually limited minds, directing attention and coordinating complex tasks. Of course, institutions may generate their own unexpected complexities, which in turn have to be managed.

Simon’s arguments can be reframed in terms of a more recent literature in complexity science on ‘coarse-grainings.’ Crudely speaking, a coarse-graining is a stripped-down representation of some complex phenomenon that seeks to capture its key aspects and dynamics. They are ubiquitous because no scientific model, organism, or artifact can actually grasp the full detail of its environment. Rather than the crawling molecular chaos of physical reality, they always deal with abstractions, compressed and selective summaries that ignore most details. In this sense, coarse-grainings include not only sophisticated mathematical models, statistical approximations of economies and turbulent weather systems, and “blurry JPEGs” of the World Wide Web (Chiang, 2023) for example, but the individual and collective representations that many social animals, including human beings, use to keep track of social structures and relationships. Macaques, for example, maintain a rough consensus over power relations within the troop through the exchange of subordination signals, which can reasonably be understood as a coarse-graining (Flack, 2017). This consensus roughly maps expectations over which macaque is capable of beating which others in a fight, allowing “individuals to make predictions without requiring they indefinitely store all of the details of their interactions,” and consequently guiding their decisions to fight or submit. Monk parakeets have cruder representations of their relationships with conspecifics (DeDeo, 2017). Humans, in contrast, can form more sophisticated coarse-grainings, even without larger institutions, thanks both to specialized mental modules (Boyer, 2018) and social conventions such as gossip (Origgi, 2017), which allow us to keep track, for example, of shifting coalitional relationships.

Yet such small-bore representations are hopelessly inadequate for modern human societies, which require impersonal social and informational technologies that can summarize social relations at very large scale. Rather than tracking a few individuals in a close-knit hunter-gatherer community where everyone knows everyone well, or even a village or town, we need to manage interactions that may involve millions—even billions—of people at once. Building and improving the means to do this has involved the development of institutions such as markets, bureaucracies, and even democracy that can handle relatively impersonal relationships at scale, using coarse-grainings that make these relationships comprehensible.

To understand how AI is reshaping society then, we can start by treating AI as a novel social technology, relying on coarse-grainings like all the others, and investigating how it interacts with existing social technologies. This integrates arguments over AI with older debates over the Long Industrial Revolution. Economic historians have documented how the transformation of human capacities to produce at scale over the last couple of centuries was facilitated by ‘general purpose technologies’ such as steam power and electricity. While many scholars have asked whether AI, too, is a general-purpose technology, they have paid less attention to the institutional aspects of the Industrial Revolution; most prominently the profound transformations in human capacity to organize economic, social, and political life, which enabled new technologies of production.

As STS scholars have consistently emphasized (Yates (1993) is particularly strong on this), bureaucratic innovations enabled the creation of new organizational structures that allowed for massive coordination. Classical liberal economists like Hayek have explored how large scale markets can compensate for the “computational limits of human beings,” Simon (1996, 35) allowing actors to make economic decisions on the basis of limited local information. Social scientists like Anderson (1991) and Gellner (1983) have debated how “print-capitalism,” the articulation of the nation state, and new forms of schooling and information dissemination made large scale publics (crudely) legible to themselves as well as their rulers, potentially stabilizing both democracy and responsive autocracy.

All these systems generate, and rely on, coarse-grainings. Bureaucratic standards and statistics summarize complex social realities and make them legible, for better and for worse (Scott, 1998). The price mechanism radically simplifies the intangible aspects of economic production, allowing market actors to devote their attention to buying and selling (Hayek, 1945). Opinion surveys, censuses, and other means simplify public opinion and make it legible to ordinary people and the political actors empowered to act on their behalf (Perrin and McFarland, 2011), and were relied on even by authoritarian states to map what their subjects wanted and thought (Dimitrov, 2023).

This vast increase in the social capacity for complex information processing has enabled collective cognition and problem solving at a historically unprecedented scale. Social technologies like markets, bureaucracy, and democracy allow human beings to become what economic historian Brad DeLong (2026) calls an “anthology intelligence,” capable of deploying accumulated cultural knowledge in a coordinated way towards large-scale ends. That is their positive aspect. In their negative, these systems regularly appear monstrous to those who find themselves at the wrong end of the power relations they create. Markets, bureaucracies, and even democracies have furthermore devoured older and more intimate forms of social organization, replacing them with vast systems that are regularly indifferent, and sometimes inimical, to the particular fates and desires of individuals and groups.

A social technology perspective then emphasizes just how long established the terms and language of today’s AI debates are. The older guiding systems of modernity—just like currently existing AI—lack anything that can reasonably be described as intentionality (Shalizi, 2010). However, humans have regularly treated bureaucracies, nations, and even markets (Spufford, 2017) as if they were individual intelligent agents, whether they had benign or hostile intent and goals. Current descriptions of AI regularly steal metaphors and ideas from these past debates, whether they draw on racist-conservative condemnations of modernity and democracy (Lovecraft, 2005; for a rejoinder, see Bear, 2013), humanist critiques (Jarrell, 1941), or anarchistic eulogies to past ways of life (Scott, 1998).

More importantly, this perspective allows us to begin comparing the actual workings of older social technologies with the newer ones, and to start mapping what happens as the two interpenetrate. AI is not the apotheosis of the robot gods nor yet of their human masters. It is a new machinery of complex information processing, perhaps even comparable to markets, bureaucracies, and democracy. Its one weird trick is to take enormous bodies of digitized information, whether social-economic, textual, visual, or otherwise, and generate abstractions that look to capture their leading statistical characteristics.

This is true across the various forms of AI that are deployed today. When social media platforms guess which post to serve or movie to recommend, they match particularized ‘embeddings’—coarse-grainings of information about the particular users and the universe of content—to arrive at their predictions. The LLMs that we emphasize are no more than coarse-grainings of the vast corpora of textual information that they have been trained on, post-processed to seem more natural in their interactions with humans and carry out more complex tasks. They are also no less. It is astonishing that we now have manipulable representations of entire bodies of human culture which can be set to work via an ordinary language interface to produce new outputs. These technologies are a new stage in the trajectory of institutional and organizational development that has run through modernity and the Long Industrial Revolution, giving new ways of managing complexity while creating their own complexities too.

Models, Lossiness and Power

Obviously, AI is not just one of the social technologies of the past. We emphasize the need for better mappings of what happens when it collides, and meshes, with its predecessors. Such mapping requires some understanding both of how the newer technology works and its consequences for social organization.

Some characteristics of actually-existing AIs arise from the fact that they are not the orderly systems of symbols, rules, and heuristics which pioneers like Simon anticipated building. Rather, they are families of messy statistical models, trained on large data sets. Training a statistical model for good on-average performance implies trading worse performance in rare situations for better performance in common situations. Such trade-offs have unfortunate implications for the capacity of all such models to handle situations raised by small groups of people, by those less well-represented in the training corpora, or indeed by anything genuinely novel. These models may not even give any signal that they are operating in such regimes (although there is a research program on uncertainty quantification in LLMs that aspires to catch this). 1 1. For a not-too-obsolete survey of this rapidly-advancing area, see Liu et al. 2025. We crudely sketch one line of work, to give the flavor. The statistical technique of “conformal prediction” (Vovk et al., 2005) replaces definite or “point” predictions with the whole range of possible predictions that “conform” to the patterns established by previous data points, i.e., all predictions which would not be low-probability outliers. The breadth of “conforming” predictions then indicates the model’s uncertainty, from the near-certainty of a narrow range around a particular point to the “anything goes” of very wide intervals. Applications of these ideas to LLMs (e.g., Lin et al. 2024; Quach et al. 2024; Mohri and Hashimoto, 2024; Cherian et al. 2024), while sometimes motivated by removing hallucinations and improving “factuality”, are really means of gauging the uncertainty of the model’s output in different situations.

LLMs, and most other modern AI systems, are in fact a particular kind of statistical model: they use “deep neural networks” (LeCun et al., 2015)—that is, multi-stage, iterative function approximation—to classify, to predict, and to generate. Training and running deep neural networks is a massive exercise in matrix algebra, calling for specialized hardware. Closer to our theme, neural networks do not simply produce coarse-grainings as outputs but employ them internally. The ‘transformer’ architecture that underlies LLMs illustrates the difficulties. Like other deep neural networks, a transformer involves layers of units (‘neurons’) which form increasingly complex representations of particular aspects of the data fed through them from beginning to end. Information about mis-predictions is then propagated back through the layers to reduce errors and move the predictions close to some rough optimum. After much expense (for the genuinely large models), the transformer produces a self-standing model of the statistical patterns in sequences of text-tokens that can, after fine-tuning and tweaking, be deployed in many ways.

While there is no mystery to the math or its implementation in silico, many details of how LLMs work remain opaque. The nature of the coarse-grainings that the transformer employs internally to capture the statistical relations is often obscure. Again, there is a cottage industry devoted to ‘interpretability,’ aiming to map these relationships, but it faces stark challenges. Some internal coarse-grainings may be orthogonal to any human concept. Others may be possible to identify broadly (e.g., this unit, or pattern of units, may be associated with this nameable aspect of the data), but without understanding exactly what is going on. (Whether any other model architecture that worked as well would be equally incomprehensible is a question for future research.) As a general matter, getting these models to work well resembles the mysteries of “alchemy” more than reproducible scientific technique (Rahimi and Recht, 2017).

Finally, there are important differences in the feedback relations between various forms of statistical-learning-based AI and the social, economic, and political phenomena they look to model. Frontier LLMs are extremely expensive to train, and so are updated only at lengthy intervals. The models used for advertising and social media matching are much cheaper to estimate, and so can be changed much more quickly. Such differences affect, for example, the time scales and hence the feedback relationships between these various coarse-grainings and the social systems that they seek to represent (Flack, 2017). Those feedback relationships may of course also respond to other changes, including technical tweaks (e.g., hidden prompts to LLMs can be quickly updated) and shifts in the underlying political economy. Broader technological changes may of course have consequences too, as new tools are developed and older tools improved. For example, a near-future in which small, cheap language models can do much of the work of big ones will encourage a less centralized political economy than if large models retain a decisive edge.

Each of these aspects of modern AI calls for its own research agenda. Engaging with the particulars of the technologies, rather than stylized accounts, makes it hard to even think straight about their immediate consequences, let alone offer confident predictions about the future. Hard or not, we do need to think clearly, since these coarse-grainings and the feedback dynamics around them are reshaping society. As DeDeo (2017, 8) wrote presciently in 2017:

The recent success of deep learning is in part due to its ability to adapt, at the same time, its method of coarse-graining and its theory of the logic of those coarse-grained variables. Once we realize that the machine-aided predictors of a system are also participants, it is natural to ask how their use of that knowledge, accurate or not, back-reacts on the society itself…We understand very little about how the introduction of these prediction algorithms, on a large scale, will lead to novel feedbacks that affect our political and social worlds; it remains an understudied and entirely open topic.

Nearly a decade later, the topic is still open. So where do we even start? We suggest emphasizing two important dimensions of comparison and interaction: lossiness and power. Coarse-grainings are lossy by definition, raising the question of exactly which information gets discarded, and which is retained. Coarse-grainings also regularly get embroiled in power relations, because abstractions can create winners and losers in many social conflicts. The two often affect each other: different simplifications will advantage different groups.

Coarse-grainings necessarily discard information to create manipulable but lossy abstractions. As Maxim Raginsky (2025) puts it, “abstraction hides a great deal of complexity from view, and this is both its main virtue and its primary peril.” The protocols that allow the internet to function at scale hide a great deal of heterogeneity, exposing it to unexpected perturbations and failure modes. Much the same is true of the simplifying statistics through which central bankers perceive the economy (Davies, 2025), reducing down the complexity of vast economic systems into a small number of target variables such as inflation that can be monitored for signs of instability, and the categories (such as census classifications) through which bureaucrats see the societies that they seek to order (Scott, 1998). All ignore some aspects of the system in order to focus attention on others. Mathematical studies of coarse-graining show that the repressed aspects of the process at best return as statistical noise (Chorin et al., 2000; Crutchfield and Feldman, 2003), if not as systematic errors. AI coarse-grainings will have different attentional trade-offs than previous social technologies, but they will have trade-offs.

Such blind spots are an inevitable feature of all abstractions, and are even plausibly useful (Naidu, 2020). Equally, social technologies do not just create blind spots but feedback effects between lossy coarse-grainings and the dynamics of the process that they seek to model and perhaps govern. The logic that Scott (1998) identifies, in which bureaucratic simplifications may reshape social organization in their image or create pushback (Scott, 1985), broadly characterizes the simplified representations of “high tech modernism” too (Farrell and Fourcade, 2023). Instead of census categories, platform companies use machine-learning generated ‘embeddings’ that reflect their best guesses about which videos we might want to see next, perhaps gradually reshaping our self-understandings (Fourcade and Healy 2025).

Since socially-relevant coarse-grainings regularly create winners and losers (Jacobs and Wallach, 2021; Jacobs, 2021), they can themselves become the topic of vigorous contention. Social technologies such as bureaucratic standards and market categories relieve the burdens of unmanageable complexity by creating summaries or otherwise focusing attention on some “aspects of the situation” (Simon, 1997, 101) rather than others, thereby affecting who gets what. Cronon (1991) provides a nice example of how coarse-graining led to struggles over actual coarse grains in nineteenth century Chicago. Creating national grain markets required defining and assessing different broad categories of grain, so that buyers could distinguish good grain from mediocre or bad without having to inspect it themselves. This facilitated trade; “all honest members benefited from knowing exactly what they were buying and selling,” (119) but the crude scheme of gradation advantaged the elevator owners who bought grain, allowing them to “mix across grades,” at the expense of the farmers they bought from (134), for example, combining just enough high quality grain with lower quality product that the final mix qualified for a more lucrative grade. These crude categories generated considerable unrest among farmers, who felt that they were being stolen from but had difficulty mobilizing against a technical-seeming system that was rigged in ways that were difficult to explain.

We can begin to compare various forms of AI to existing social technologies (and to each other) by examining the specific ways in which they are lossy and affect power relations. This also helps us to begin to understand how new and old social technologies will affect each other as they inter-mesh. True understanding will require real research efforts across disciplines that talk to each other much less than they ought, such as computer science, political science, political theory, public administration, sociology, law, communications, and science and technology studies. In the interim, we can at least start talking about what such efforts could study.

AI and Bureaucracy

One urgent topic is the relationship between AI and bureaucracy, where there is already debate—some of it resting on the myths with which we began.2 2. But only some. There is a broader emerging literature on AI and bureaucracy (e.g. Caplan and boyd 2018, Fourcade and Gordon 2020, Burrell and Fourcade 2021, Cuéllar and Huq 2022) which we don’t try to summarize, instead concentrating on a narrower line of argument. Although bureaucratic hierarchy is widely disparaged for its inefficiencies, it is an essential underpinning of large-scale human civilization. Hostility to bureaucracy primes audiences for stories of how preternaturally effective AI will soon scour away the imperfections of human administration, replacing them with algorithms. Such ideas reflect the disdain that some Silicon Valley elites have for human judgment and expertise and have provided real intellectual support for the Trump administration’s program to hack away the roots of the administrative state. They are also bad starting points, even apart from their political uses. Instead of myth-making about what a stylized future technology might do to stereotyped bureaucracies, we should ask how new and ineluctably messy social technologies such as AI combine lossiness and power relations in different ways than older and ineluctably messy social technologies such as bureaucracy, and what happens when the two technologies become entangled.

AGI-centric accounts depict bureaucracy as centralized coordination, the key problem being how lower layers fail to implement the priorities of the top. Leaders’ edicts get distorted or lost as they filter through the layers of administration. AGI then figures as a general solution, replacing imperfect human bureaucrats with efficient AI agents or interfaces that will do just what they are told to do. Such accounts suggest AGI will replace bureaucratic distortion with (relatively) lossless and efficient algorithmic decision-making and implementation. That will empower those at the top, whose decisions will finally be executed just as they want them to be.

These stories make it easy to see the attractions of AI for autocratic rulers, for example. AI systems sound more efficient, and more trustworthy than human bureaucrats. Some have indeed claimed (Harari, 2018) that AI supercharges dictatorships and undermines democracy by facilitating manipulation at scale. Speculative arguments about AGI helped inspire Elon Musk’s DOGE project, which sought to hack away great swathes of America’s administrative machinery. Much of DOGE’s work and aspirations involved the application of LLMs and AI ‘agents’ to accomplish a variety of open-ended tasks (Gilbert and Elliott, 2025; Haskins and Elliott, 2025), often very badly. Shortly after Trump’s election to a second term, one of us asked a DOGE-adjacent individual whether firing multitudes of expert employees would be a problem for U.S. national security. The reply was that this was a non-issue, since AGI would happen in 2026. The government would have to eliminate most of these positions then anyway.

Bullock et al. (2025) make a less ideologically loaded version of the case that AGI will transform bureaucracy in these ways. It, too, explicitly builds on Simon’s arguments, but it claims that AGI will largely solve the problems of limited human capacity that Simon identifies.

The authors claim that we are on the verge of AGI that will match or exceed human decision-making across all domains, allowing the replacement of human beings with AI systems that enhance hierarchical power. Specifically, they say that Simon’s concerns might be addressed through the construction of a “single, vast, AGI system, similar to the general multi-modal frontier systems, in which a single interface can be deployed to effectively, efficiently, and dispassionately complete all required tasks” (25). A more Weberian understanding of the problem might alternatively task individual AGI agents to use “clearly identifiable objective factors, logic, and statistical-based, dispassionate reasoning to carefully weigh trade-offs across alternative choices” (25).

The first option would open the bottlenecks of bureaucratic attention via subsystems capable of deploying parallel problem solving that would “dynamically alter the flow of needed information and decisions to accomplish each task most effectively, and at an inhuman speed” and be “deployable to perform any arbitrary set of tasks,” while the second would replace the problems of “human dominated bureaucracies ... littered with subjectivity and emotion-centered decision making,” with “a high-level of predictability across situations with similar contexts, leading to overall improvements in the predictability of bureaucratic decision making” (25). The authors accept that such increased efficiencies would come at a cost. An AI-empowered state could be associated with increased inscrutability, misalignment from broader social goals, and reduced prospects for individual liberty and freedom.

Such ideas have provided intellectual justification for what one of the paper’s authors enthusiastically describes elsewhere (Hammond, 2025) as “DOGEmaxing”: enabling “software to eat the state” while fostering the “crowding-in [of] privatized forms of governance.” As the primary drafter of the Trump administration’s AI Action Plan argued shortly before he took up his White House position (Ball, 2025):

Today, when the CEO of a company wants to make some change to a business process, they relay that command through chains of leadership, and each time it loses some fidelity …[Now] CEOs and managers will be able to say “jump,” and in unison, tens, hundreds, thousands, or millions, of agents will say “how high?” I note, with interest, the fact that this technology is being built at the very time that the Republican Party, and Donald Trump in particular, seek to advance theories of a “unitary executive”—the notion that the President exercises the powers granted to him by the Constitution and by Congress absolutely. By the end of President Trump’s term, that may be more possible than anyone ever imagined.

While Ball’s politics are more complex than this might suggest (Klein, 2026), he reportedly (Bronzini-Vender, 2026) continues to argue that AI:

might replace the Supreme Court, then the United States government itself. AI, [Ball] said, was “this giant acid vat” dissolving society’s mediating institutions. “Future institutions will be machinic,” he said. “It will not be AI in government. It’s going to be AI as governments.”

We certainly can’t blame AGI speculation alone for DOGEmaxing and the Trump administration’s evisceration of the federal bureaucracy; there are many overlapping causes. However, if you believe that the duty of a bureaucracy is to implement the leader’s program, and AGI is nigh, it is not hard to conclude that the latter offers a providential way to accomplish the former. After all, many other Silicon Valley innovations have replaced purportedly messily inefficient social systems (ranging from regulated taxis through employment relations for lower-level Amazon workers to general forums of public debate) with optimization of a function that weights the leadership’s goals and objectives, and algorithmic feedback loops that implement it.

As Narayanan (2026) says, discussing the application of closely related ideas to radically different ends, “almost everything about this naive mental model is wrong.” Bureaucracy involves far more than the automatable implementation of a pre-cooked program, and neither frontier models nor their near-future descendants will be systematically better at reliably making bureaucratic trade-offs than collective human processes. Partly this is because of problems with metrics. Optimization requires quantified metrics, which may be poorly matched to underlying concepts and actual goals. Nguyen (2024) presents the example of Netflix researchers trying to train generative systems to produce “good art” by optimizing on the number of engagement hours. It may actually be more difficult to ‘optimize bureaucracy towards Trump’s end-goals,’ given how chaotic and changeable they are, than to optimize towards good art. There are also limitations specific to modern AI. The LLMs at the core of multimodal models are not, in fact, optimizers. (Their parameters are adjusted by optimizers during training, but that’s a different thing). They tend increasingly to become “hot messes” as they continue to try to reason, in ways that cannot readily be solved by scaling (Hägele et al., 2026).

However, the most basic objections to this vision stem from the nature of optimization itself. Contrary to the language about “carefully weigh[ing] tradeoffs,” it simply does not provide any objective means of weighing the kinds of choices across non-commensurables that are essential to the bureaucratic process. Simon, Smithburg and Thompson (1958, 74) discusses how both the Work Projects Administration and the War Production Board had to choose policies across widely divergent goals in the first half of the twentieth century; “pump priming” versus immediate help to the unemployed in the former case, and finding tradeoffs between war goals and civilian needs in the latter. As they note, such “internal conflicts and contradictions among the ultimate objectives, or among the means selected to attain them” imply that “the means-end hierarchy is seldom an integrated, completely connected chain.” Such vexing choices regularly emerge in bureaucratic implementation and articulation, as well as in goal-setting at the top, making them fundamentally political problems, and this is just what optimization does not solve. As Recht (2023) bluntly says, “you can’t optimize a trade-off.” There is no consensus over how to optimize over multiple independent objectives in even moderately complex situations. Machine learning, with or without neural nets, does not offer magical solutions to this. (Cf. McCulloch, 1945.) Of course, it is possible to ‘ask’ LLMs to balance different goals against each other and then look at the outputs, but there is no generalizable path towards measuring their success, let alone improving it across repeated iterations. Invocations of ‘AI’—let alone ‘AGI’—ignores the possibilities and limitations of these technologies for bureaucratic purposes, substituting magic in their stead.

Again, Simon insistently and repeatedly emphasizes that bureaucratic decision-making involves trade-offs between hard-to-compare objectives, and repeated reformulation of both goals and means of achieving goals, at all levels of the organization as people try to satisfice across the complexities. Lindblom concludes that such continual up, down, and sideways adjustments are why we should not think of bureaucratic coordination as just a top-down process:

The task of coordination is often identified with that part of it which is in fact attacked through central coordination; what comes clear on second thought is forgotten—that an enormous amount of coordination is inevitably achieved through various mutual adjustments. But the most visible part of the coordination iceberg, explicit central coordination, may be only a small part of all those processes through which coordination to a degree is achieved (Lindblom, 1965, 170).

Implementing top-down orders is but one (important) part of bureaucracy’s work, which requires repeated mutual adjustments of a sort that resist standard AI approaches (Narayanan, 2026).

So how might we start doing justice to the ideas of Simon, Lindblom, and others who have thought carefully about how bureaucracy works? We must acknowledge that there are important problems of central coordination, but AI coarse-grainings carry their own difficulties there. For example, the party officials who lead the People’s Republic of China are regularly frustrated by their inability to know what is happening below them. Lower-level officials regularly distort policies to their personal advantage, and have also squeezed off alternative flows of information to the top, by, for example, forcibly discouraging the public from complaining to central Party officials (Anderlini, 2009). As Wallace (2016) demonstrates, higher-level Chinese officials have thus employed official statistics as crude metrics of policy success, rewarding or sidelining provincial administrators based, for example, on whether GDP has increased in their province. This, unsurprisingly, has inspired lower-level bureaucrats to figure out ways to “juke the stats,” and corresponding counter-efforts by their superiors to find more reliable metrics.

All this seems to suggest that LLMs and other forms of modern AI will increase the power of central officials, refining their control of their subordinates. And that might happen! Yet if senior party officials turn to LLM-generated reports as assessments of junior officials, the cat-and-mouse games will continue in a new medium. Already, people in other spaces are trying to juke the training data to push LLMs in one direction or another. Officials whose careers are at stake, and who can direct underlings to write internal documents or otherwise spin the LLM’s sources, may have means of making themselves look good. Whether LLMs will give senior officials an advantage over their subordinates, or vice-versa, will involve details of the technology, its application, and human responses. It is likely to result in new trade-offs, and new competitive relations between affected human beings, rather than the sort of radical transformation that can be posited through ex ante speculation.

Similarly, no one really knows whether powerful AI will make society dramatically more vulnerable to authoritarianism, reshaping power relations between government and citizens as Bullock et al. (2025) suggest. It is not just (as Bullock et al. acknowledge) that non-state actors may themselves use technology to obfuscate the data, but that the data may not be all that useful to start with. Systematically incomplete data is liable to worsen bureaucratic blind spots rather than compensate for them. Most pertinently, as Yang (2026) argues, authoritarian states face profound trade-offs in deploying AI to detect dissidence. Their institutions repress the overt behavior that might generate training data to predict future unrest. While AI certainly increases the power of authoritarians to monitor citizens at scale, it also plausibly increases their reliance on coarse-grainings that may not capture what they want to know most, biasing them towards predictions based on past data that may be unsuited to future contingencies (Anastasopoulos and Lian, 2026).

Both authoritarian and non-authoritarian states also face ideological trade-offs. LLMs are extremely well suited to the articulation, promulgation, and interpretation of organizational ritual (Fourcade and Farrell, 2024), such as writing organizational boilerplate language, explaining why your activities contribute to the goals of the larger organization, and similar. Rituals may play an important coordinating role within organizations and elsewhere (Chwe, 2013). Simon stresses the importance of very general summary representations of shared goals in building organizational culture. Such abstractions provide generic guidance about how to apply broad priorities to particular problems. Here, LLMs might allow such representations to speak and answer questions, creating ideological oracles that can articulate how general goals relate to specific situations, including ones never anticipated by the original drafters. An LLM might, for example, explain—plausibly and in detail—how to apply China’s ‘dual circulation’ strategy to the (important) ball-bearing manufacture sector, even if senior Party officials had never actually written particular guidance on the topic. The current U.S. Undersecretary of Defense for Research and Engineering wants Pentagon employees to use AI agents to check “proposed actions against the tenets of the national defense strategy” (Manson, 2026).

Equally, such coarse-grainings may create systematic blind spots around ideologically inconvenient facts or questions. Chinese LLMs such as DeepSeek are trained to avoid direct answers on topics such as the Tiananmen Square massacre (Lu, 2025). This training is spotty, but may likely improve—‘constitutional’ AI techniques (Anthropic, 2026) can surely be deployed for such ends. Yet if LLMs become load bearing parts of the ideological infrastructure, such blind spots may have negative consequences as well as benefits, making it harder for officials to see the fissures that they conceal by seamlessly stitching other facts and ideas together. There are historical precedents. In Maoist China, provincial statistics systematically exaggerated the size of harvests, leading central authorities to take the crops, leaving over 30 million people to starve to death (Wallace, 2014). As a general matter, higher ideological coherence and coordination power may come at the expense of increased lossiness and inability to see problems whose nature is at odds with the regime’s organizing myths. While there are possible fixes, there are some reasons to suspect that increased use of AI may worsen rather than alleviate such problems. Other interpretations of the technology suggest countervailing forces to the tendency to conformity, but these may have their own trade-offs. The potential coordination value of LLM based ideological-cultural oracles is partly offset by the risk that these oracles may equivocate, or, even worse, provide radically different responses to different questioners, based on differences in prompts—exact word choices, implicit assumptions, or artful manipulation by bureaucrats with their own agendas. The extent of this problem will depend not only on the technology itself, but the institutional circumstances in which it is used (who has access and under what conditions).

To the extent that it is possible to eliminate variability through fine-tuning, or just through controlling the circumstances under which LLMs are used, other trade-offs will again arise. Existing inefficiencies in the promulgation of top-down directives actually have advantages. Lower-level bureaucrats may understand local or particular circumstances that the top of the hierarchy does not. Providing these bureaucrats with leeway to adapt policies to such circumstances may risk coordination failure or malfeasance, but may also allow better overall outcomes than would a combination of general-purpose abstractions and centralized micro-management.

Existing bureaucratic structures face trade-offs in horizontal as well as vertical coordination and information sharing. LLMs are remarkably well fitted to mitigate some of these trade-offs, but not others. Some of these uses may appear trivial but are in fact highly important.

For example, discussions with former and current military officials suggest that the US armed services are plagued by differences in terminology which may lead to serious coordination failures when different branches look to work together. The problem is so serious that the Department of Defense has a dedicated Terminology Program (Joint Chiefs of Staff, undated), reporting to the Joint Chiefs of Staff, which seeks to press back against idiosyncratic language used by different branches of the armed services by creating commonly agreed terms and definitions. LLMs are extraordinarily well suited to mitigate such problems by translating common objectives or requirements across different branches of an organization that have their own jargons. Their compressions omit some details, but seem in general to be highly proficient at capturing genres, including bureaucratic genres, and translating across them. That may allow the automation of many routine tasks of organizational translation that currently require sustained human effort. Equally, automated translation may carry its own hidden burdens. Some of the frictions of translation represent genuine operational differences in how different branches carry out their work in the real world, and over-facile translation may obscure these differences—until bureaucratic language meets material reality. On average, the benefits likely greatly outweigh the costs, but there are costs.

Such problems will likely intersect with power struggles and clashing political interests. The U.S. armed services have their own internal rivalries, sometimes encouraged by the politicians who oversee them. Overlap in needs and expertise across the different branches may encourage tale-telling: if an Air Force proposal is an expensive vanity project that sucks up more of the defense budget than it should, then the U.S. Navy, which also has planes, may have both the technical understanding and the incentive to inform its friends in Congress about the problem. We don’t pretend to predict how easier coordination across the armed services will intersect with clashing interests between them, but we do expect that it will matter. Similar conflicts lurk within all complex organizations, and self-interested individuals and factions are remarkably adept at turning better coordinating technologies to their own private purposes. So too, incumbent market actors will seek enthusiastically to use technologies to deepen moats rather than drain them (Curl, Kapoor and Narayanan, 2026).

Democratic feedback faces related problems. There is much enthusiasm about the potential of LLMs, for example, to make bottom-up public feedback to bureaucratic decision-making less lossy than it is. Algorithmic feedback loops could ensure that problems not only become visible more quickly, but are sometimes resolved automatically in the implementation stage. LLMs could make public comment processes legible in new ways, summarizing and remixing public feedback at scale.

Such innovations are worth pursuing but face sharp limits. Bureaucratic processes are notoriously terrible at incorporating feedback, but AI approaches may actually be worse at incorporating unexpected information. The “street level algorithms” that actually implement decisions may well do worse than human bureaucrats at identifying and responding to anomalous events (Alkhatib and Bernstein, 2019).

When street-level bureaucrats encounter a novel or marginal case, they use that case to refine their understanding of the policy. When street-level algorithms encounter a novel or marginal case, they execute their pre-trained classification boundary, potentially with erroneously high confidence.

Such algorithms may not even be able to see the ways in which a specific case has surprising features that do not fit its patterns. Nor is this fundamentally different for LLMs, or for that matter as-yet-to-be-invented-forms of AI. Again, all such statistical models are optimized for good-on-average performance rather than dealing with rare events that seem anomalous. Different implementations will face different versions of this dilemma, but all will face it. Berliner (2024) describes how similar problems plague the automated analysis of public comments, where LLMs, like other forms of natural language processing, are strong at identifying and clustering broad patterns, but weak at detecting rare, unexpected, and for just that reason, valuable information. In his words:

Much of the promise of public participation in government is precisely its ability to inform policymakers about novel problems, novel solutions, or novel perspectives from the public.

However, AI is poorly suited to identifying novelty:

If [policy makers] want to be able to learn things that are both specific and novel—where information processing requires both individual attention and context-specific knowledge—then AI is not likely to be very useful unless society is willing to accept either serious biases or major inaccuracies. And restricting the scope of policy learning to already known measures and categories could even serve to simply lock in existing political inequalities.

As Berliner notes, when human decision-makers rely on algorithms that are good at capturing certain kinds of information, it may actually exacerbate such systematic lossiness, potentially endangering the organization’s capacity to respond to unexpected circumstances. AI reductions of public understandings make visible patterns that were previously too vast to easily discern, but at the cost of hiding details that do not fit the models’ abstractions.

Such trade-offs also appear in more transformative use-cases. LLMs allow for new syntheses of information, which, while still lossy, will capture aspects of organizational knowledge that other technologies cannot. Many, though not all, of the problems that scholars like Hayek and Lindblom associate with the ‘synoptic’ vision of the state stem from the differences between formal bureaucracies and more dynamic social technologies such as markets. As many have pointed out, large organizations do not know what they know: enormously valuable information may be scattered across the organization, and resist being assimilated and analyzed in one place. The paperwork in filing cabinets in a field office in Nebraska might turn out to have valuable implications if combined with statistical data gathered in Florida and centrally collected national data in the suburbs of Washington D.C.

Once such resources are made machine-readable, LLM-based systems can go a remarkable way towards stitching disparate sources of information together to make it more useful, ameliorating Hayek’s and Lindblom’s synoptic problem (Brynjolffson and Hitzig 2025). This is potentially extraordinary, for better or worse. Enthusiasts see it as a radical empowering of the state to do better things for its citizens. Skeptics fear that this will eliminate many freedoms that depend on the continued existence of imperfectly visible spaces where the state is not able to pull together everything that it knows (Fourcade and Healy, 2025). Military AI may not so much replace bureaucracy as they “encode” it into models that, like many of their paper-based predecessors, “achieve completeness by filtering out everything that wasn’t legible to their categories,” and treat deliberation over agonizing choices as inefficient latency that ought be eliminated (Baker, 2026).

Such models do not create the “panopticon” that utilitarians celebrated and critics of state power like Foucault (1977) feared, but structures that resemble it in many significant aspects. One might describe the result as a mosaic eye synopticon, potentially allowing bureaucrats at all levels to pull potentially useful information from across the organization, according to specific prompts, and towards specific purposes. Equally, it is a cockeyed synopticon. Everything it stitches together will itself be a coarse-graining of a more complex reality, while the end product will be its own crude piece of stitchwork, throwing away such of the fine-grained data as does not meet its own criteria. The end results will be imperfect in their own particular way. The worst-case scenarios might resemble the all-seeing machine state of Orwell’s 1984 less than Terry Gilliam’s Brazil, in which an insect jammed in a teleprinter creates a series of self-compounding errors. The best-case outcomes will still have a lot of decisional slop mixed in with broadly benign policy outcomes. There will be an enormous variety of possibilities in between the two. Again, we don’t offer strong predictions about outcomes, but we do know that these technologies will be lossy too, albeit in different ways than their predecessors, and that different factions will struggle over how to ensure that the associated blind spots and advantages of coordination are turned to their own benefit, rather than that of rivals.

To sum up: AGI does not hold out the promise of truly post-human bureaucracy. Many bureaucratic processes (including ones that we have not talked about) could be made more efficient, some perhaps even completely automated by AI. But AI—including any plausible version of AGI descended from current technologies—cannot replace bureaucracy in the ways that enthusiasts promise. The frustrations of actually existing bureaucracy do not merely arise from inept or technically-inadequate solutions to the principal-agent problem. They emerge too from the collision of multiple incommensurable demands, each with its own problems and benefits, so that there are no optimal design solutions. Those who build or reform bureaucracies, like those who build other artifacts, need to satisfice across multiple intersecting needs and pathologies. Designs that neatly address one kind of problem may radically worsen others. Actually-existing AI has its own imperfections, some of which are endemic. Grafting AI systems onto existing bureaucracies will solve some problems but will worsen others and make altogether new ones. It will not eliminate the political difficulties of mediating across different, often non-commensurable, goals. Imagining replacing bureaucracy wholesale with AI is only plausible if one waves away the actual difficulties associated with real social technologies.

Our own hesitant bet is that LLM-empowered bureaucracy will be less amenable to control than critics fear and proponents hope. Current LLMs are architecturally opaque, in part because it’s very obscure how they coarse-grain their inputs. This makes it harder to deliberately tilt them in various directions, at least not with any precision, but it also raises issues for bureaucracy. The (practical) impossibility of knowing which aspects of incoming signals are filtered out by the apparatus makes it especially hard to compensate for the resulting information loss.

This may however mean that chimerical melds of bureaucracy and AI will be subject to more contention rather than less, and that the contention will get weirder. The aspects of the signals lost to compression might be extremely hard to describe in previously-existing humanly-meaningful terms, but they might well make more trouble for some people than others nonetheless. Just as nineteenth century Midwestern farmers had difficulty in articulating how exactly they were hurt by grain classification schemes, twenty-first century citizens and interest groups may flounder while describing what AI is doing to them, even if they know that things are happening that they do not like.

Very large-scale transformations in our political economy may already be underway, but may be extremely difficult to explain in the traditional terms that we have used to make sense of political conflicts between different classes of actors with different interests. That the terms of such struggles are unfamiliar or difficult to articulate in our current language do not mean that we ought to ignore or discount them. Similar difficulties have attended previous major economic disruptions in the Long Industrial Revolution. Just as we should rescue past ideologies from the “enormous condescension of posterity” (Thompson, 1963), so too we need to build a more serious and systematic understanding to help us, and other people too, understand and articulate what is happening right now. Properly mapping the political economy of AI and bureaucracy, then, would be one part of a broader project to understand the political economy of AI.

Conclusions

We are not setting out to test AGI-centric accounts against the social technology approach in some purportedly scientific way. That would load the dice heavily in our favor: complex description is definitionally more likely to be accurate than stylized argument, although often far less useful (Healy, 2017). Nor are we trying to predict the future. Instead, we are drawing attention to the messy, consequential interactions between actually-existing human institutions (especially bureaucracy), and actually-existing AI, that AGI-centric accounts pass over. Perhaps these will all become irrelevant when the bots come marching in, but that is a pious (or blasphemous) hope, not a rational forecast.

AI coarse-grainings—even if they are built by something that loosely approximates AGI (itself a loose notion in urgent search of an empirical approximation)—will still be lossy. (We repeat yet again, all statistical models face trade-offs between better on-average performance and worse performance in unusual situations.) The interesting questions involve the interaction between the ways bureaucracies abstract reality and the coarse-grainings that new AI applications will lead to. When will one system compensate for the deficiencies of the other? When will their different flavors of lossiness prove mutually reinforcing? What new problems may result from combining very different systems for managing complexity that are themselves highly complex? How will power relations change as a result? Who will benefit, and who will be hurt? These and other questions might be asked, pari passu about the relationship between AI and other social technologies such as markets and democracy too. We absolutely ought to start asking them.

We hope to move debate decisively away from the question of whether AI will precipitate a profound transformation in the human condition (Kokotajlo et al., 2025). AI is better understood through past history and current engagement with its actual consequences for contemporary human society than guesses about how it might work magic in the future. More grounded approaches to AI keep on getting dragged back into irresolvable fights about the exact timeline of the countdown to infinity (Kapoor et al., 2025), impeding engagement with social science about what is actually happening to human society and institutions right now. That has to change.

References

Alkhatib, Ali and Michael Bernstein (2019). “Street-level Algorithms: A Theory at the Gaps Between Policy and Decisions.” In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems [CHI ’19] (Stephen Brewster and Geraldine Fitzpatrick and Anna Cox and Vassilis Kostakos, eds.). New York: Association for Com-puting Machinery. URL https://hci.stanford.edu/publications/ 2019/streetlevelalgorithms/streetlevelalgorithms-chi2019.pdf. doi:10.1145/3290605.3300760.

Anastasopoulos, L. Jason and Jie (Jason) Lian (2026). “The Limits of Authoritarian AI,” Journal of Democracy, 37:5-17.

Anderlini, Jamil (2009). “Punished Supplicants,” Financial Times, March 6 2009. https://www.ft.com/content/7d13197e-09bc-11de-add8-0000779fd2ac. Accessed 04-17-2026.

Anderson, Benedict (1991). Imagined Communities: Reflections on the Origin and Spread of Nationalism. London: Verso, revised edn.

Andreessen, Marc (2023). “The Techno-Optimist Manifesto.” Weblog post, Andreessen Horowitz. URL https://a16z.com/ the-techno-optimist-manifesto/.

Anthropic (Jan. 2026). “Claude’s New Constitution.” URL https://www. anthropic.com/news/claude-new-constitution. Accessed: 2026-02-08.

Baker, Kevin (2026). “Kill Chain.” Artificial Bureaucracy. URL https://artificialbureaucracy.substack.com/p/kill-chain.

Ball, Dean W. (Mar. 2025). “Where We Are Headed.” Hyperdimensional . URL https://www.hyperdimensional.co/p/where-we-are-headed. Accessed: 2026-02-08.

Banks, Iain M. (1987). Consider Phlebas. London: Macmillan.

Bear, Elizabeth (2013). Shoggoths in Bloom. Prime Books.

Becker, Adam (2025). More Everything Forever: AI Overlords, Space Empires, and Silicon Valley’s Crusade to Control the Fate of Humanity. New York: Basic Books.

Berliner, Daniel (Nov. 2024). “What AI Can’t Do for Democracy.” Boston Review . URL https://www.bostonreview.net/articles/ what-ai-cant-do-for-democracy/. Accessed: 2026-02-08.

Boyer, Pascal (2018). Minds Make Societies: How Cognition Explains the World Humans Create. New Haven, Connecticut: Yale University Press. doi:10.12987/9780300235173.

Bronzini-Vender, Alex (2026). “Palantir Comes to Campus: At a Quiet Conference at Yale, the Company and Its Allies Sketched a Vision for AI, State Power, and How to Mix the Two,” New York Magazine, May 1, 2026, https://nymag.com/intelligencer/article/palantir-yale-conference-ai.html . Accessed 05-02-2026.

Brynjolffson, Erik and Zoë Hitzig (2025). AI’s Use of Knowledge in Society. National Bureau of Economic Research.

Bullock, Justin B., Samuel Hammond and Seb Krier (2025). “AGI, governments, and free societies.” In The Art of Digital Governance: Navigating Platforms and AI Revolution, pp. 327–361. Cham, Switzerland: Springer. URL https://arxiv.org/abs/2503.05710. doi:10.1007/978-3-032-00514-417.

Cherian, John J., Isaac Gibbs and Emmanuel J. Candès (2024). “Large Language Model Validity Via Enhanced Conformal Prediction Methods", pp. 114812--114842 in A. Globerson, L. Mackey, D. Belgrave, A. Fan, U. Paquet, J. Tomczak and C. Zhang (eds.), Advances in Neural Information Processing Systems 37 [NeurIPS 2024] (Red Hook, New York: Curran Associates), doi:10.52202/079017-3645, https://arxiv.org/abs/2406.09714

Chorin, Alexandre J., Ole H. Hald and Raz Kupferman (2000). “Optimal Prediction and the Mori-Zwanzig Representation of Irreversible Processes.” Proceedings of the National Academy of Sciences (USA), 97: 2968–2973. URL http://www.ma.huji.ac.il/~razk/Publications/PDF/CHK00.pdf. doi:10.1073/pnas.97.7.2968.

Chwe, Michael Suk-Young (2013). Rational Ritual: Culture, Coordination, and Common Knowledge. Princeton, New Jersey: Princeton University Press. doi:10.1515/9781400846436.

Cronon, William (1991). Nature’s Metropolis: Chicago and the Great West. New York: W. W. Norton and Company.

Crutchfield, James P. and David P. Feldman (2003). “Regularities Unseen, Randomness Observed: Levels of Entropy Convergence.” Chaos, 13: 25–54. URL http://arxiv.org/abs/cond-mat/0102181. doi:10.1063/1.1530990.

Curl, Justin, Sayash Kapoor and Arvind Narayanan (2026), “AI Won’t Automatically Make Legal Services Cheaper,” Lawfare, https://www.lawfaremedia.org/article/ai-won-t-automatically-make-legal-services-cheaper.

Davies, Dan (2025). The Unaccountability Machine: Why Big Systems Make Terrible Decisions — and How the World Lost Its Mind. Chicago: University of Chicago Press.

DeDeo, Simon (2017). “Major Transitions in Political Order.” In From Matter to Life: Information and Causality (Sara Imari Walker and Paul C. W. Davies and George F. R. Ellis, eds.), pp. 393–428. Cambridge, United Kingdom: Cambridge University Press. URL https://arxiv.org/abs/1512.03419. doi:10.1017/9781316584200.

DeLong, J. Bradford (2026). “Theories of Economic History.” Sir John Hicks Memorial Lecture, All Souls College, Oxford University. April 28.

Dimitrov, Martin K. (2023). Dictatorship and Information: Authoritarian Regime Resilience in Communist Europe and China. Oxford, England: Ox-ford University Press. doi:10.1093/oso/9780197672921.001.0001.

Emrys, Ruthanna (2022). A Half-Built Garden. New York: Tor Books.

Farrell, Henry (2025). “AI as Governance.” Annual Review of Political Science, 28: 375–392. doi:10.1146/annurev-polisci-040723-013245.

Farrell, Henry and Marion Fourcade (2023). “The moral economy of high-tech modernism.” Daedalus, 152(1): 225–235. URL https://www.amacad.org/sites/default/files/publication/downloads/ Wi23_Daedalus_28_Farrell-Fourcade.pdf. doi:10.1162/daeda01982.

Farrell, Henry, Alison Gopnik, Cosma Rohilla Shalizi and James Evans (2025). “Large AI models are cultural and social technologies.” Science, 387: 1153–1156.

Farrell, Henry and Hahrie Han, AI and Democratic Publics, Knight First Amendment Institute, 2025.

Flack, Jessica C. (2017). “Coarse-Graining as a Downward Causation Mechanism.” Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 375(2109): 20160338. doi:10.1098/rsta.2016.0338.

Foucault, Michel (1977). Discipline and Punish: The Birth of the Prison. New York: Pantheon Books. Originally published in French as Surveiller et punir in 1975.

Fourcade, Marion and Henry Farrell (Sep. 2024). “Large Language Models Will Upend Human Rituals.” The Economist . URL https://www.economist.com/by-invitation/2024/09/04/ large-language-models-will-upend-human-rituals. Accessed: 2026-02-08.

Fourcade, Marion and Jeffrey Gordon (2020). “Learning Like a State: Statecraft in the Digital Age.” Journal of Law and Political Economy 78: 78-108.

Fourcade, Marion and Kieran Healy (Aug. 2025). “The Sovereign Individual and the Paradox of the Digital Age.” Aeon. URL https://aeon.co/essays/ the-sovereign-individual-and-the-paradox-of-the-digital-age. Accessed: 2026-02-08.

Gellner, Ernest (1983). Nations and Nationalism. Ithaca, New York: Cornell University Press.

Gilbert, David and Vittoria Elliott (apr 2025). “DOGE Put a College Student in Charge of Using AI to Rewrite Regulations.” Wired . URL https://www.wired.com/story/ doge-college-student-ai-rewrite-regulations-deregulation/. Accessed: 2026-02-08.

Grosse, Roger, Juhan Bae, Cem Anil, Nelson Elhage, Alex Tamkin, Amirhossein Tajdini, Benoit Steiner, Dustin Li, Esin Durmus, Ethan Perez, Evan Hubinger, Kamile Lukosiute, Karina Nguyen, Nicholas Joseph, Sam McCandlish, Jared Kaplan and Samuel R. Bowman (2023). “Studying Large Language Model Generalization with Influence Functions.” E-print, arxiv:2308.03296. URL http://arxiv.org/abs/2308.03296.

Gudiño, Jairo F., Umberto Grandi and César Hidalgo (2024). “Large Language Models (LLMs) as Agents for Augmented Democracy.” Philosophical Transactions A, 382(2285): 20240100. URL https://arxiv.org/abs/2405.03452. doi:10.1098/rsta.2024.0100.

Hägele, Alexander, Aryo Pradipta Gema, Henry Sleight, Ethan Perez and Jascha Sohl-Dickstein (2026). “The Hot Mess of AI: How Does Misalignment Scale with Model Intelligence and Task Complexity?” In Fourteenth International Conference on Learning Representations [ICLR 2026] , p. forthcoming. URL https://arxiv.org/abs/2601.23045.

Hammond, Lewis, Alan Chan, Jesse Clifton, Jason Hoelscher-Obermaier, Akbir Khan, Euan McLean, Chandler Smith, Wolfram Barfuss, Jakob Foerster, Tomáš Gavenčiak, The Anh Han, Edward Hughes, Vojtěch Kovařík, Jan Kulveit, Joel Z. Leibo, Caspar Oesterheld, Christian Schroeder de Witt, Nisarg Shah, Michael Wellman, Paolo Bova, Theodor Cimpeanu, Carson Ezell, Quentin Feuillade-Montixi, Matija Franklin, Esben Kran, Igor Krawczuk, Max Lamparth, Niklas Lauffer, Alexander Meinke, Sumeet Motwani, Anka Reuel, Vincent Conitzer, Michael Dennis, Iason Gabriel, Adam Gleave, Gillian Hadfield, Nika Haghtalab, Atoosa Kasirzadeh, Sébastien Krier, Kate Larson, Joel Lehman, David C. Parkes, Georgios Piliouras and Iyad Rahwan (2025). “Multi-agent risks from advanced AI.” E-print, arXiv:2502.14143. URL http://arxiv.org/abs/2502.14143

Hall, Andy (2026), “Building Political Superintelligence,” Free Systems, March 26, 2026.

Hammond, Samuel (jan 2025). “DOGEmaxing.” Weblog post, Second Best. URL https://www.secondbest.ca/p/dogemaxing. Accessed: 2026-02-08.

Hao, Karen (2025). Empire of AI: Dreams and Nightmares in Sam Altman’s OpenAI . New York: Penguin.

Harari, Yuval Noah (2018). “Why Technology Favors Tyranny.” The Atlantic, 322(3): 64–73.

Haskins, Caroline and Vittoria Elliott (May 2025). “A DOGE Recruiter Is Staffing a Project to Deploy AI Agents Across the US Government.” Wired . URL https://www.wired.com/story/ doge-recruiter-ai-agents-palantir-clown-emoji/. Accessed: 2026-02-08.

Hayek, Friedrich A. (1945). “The Use of Knowledge in Society.” American Economic Review, 35: 519–530. URL http://www.econlib.org/Library/ Essays/hykKnw1.html. Reprinted in (Hayek, 1948, pp. 77–91).

 — (1948). Individualism and Economic Order . Chicago: University of Chicago Press.

Healy, Kieran (2017). “Fuck Nuance.” Sociological Theory, 35: 118–127. doi:10.1177/0735275117709046.

Jacobs, Abigail Z, and Hanna Wallach (2021), “Measurements as Fairness,” FAccT 2021.

Jacobs, Abigail Z (2021). Measurement as Governance in and for Responsible AI. https://arxiv.org/abs/2109.05658.

Joint Chiefs of Staff (undated). DoD Terminology Program. https://www.jcs.mil/doctrine/dod-terminology-program/ Accessed April 17, 2026.

Jarrell, Randall (1941). “Kafka’s Tragi-Comedy.” Kenyon Review , 3(1): 116–119.

Kapoor, Sayash, Arvind Narayanan, Daniel Kokotajlo, Eli Lifland and Thomas Larsen (November 2025). “Common Ground between AI 2027 and AI as Normal Technology.” Asterisk Magazine. URL https://asteriskmag. substack.com/p/common-ground-between-ai-2027-and.

Klein, Ezra (Mar. 2026). “Why the Pentagon Wants to Destroy Anthropic. Interview with Dean Ball.” The Ezra Klein Show, The New York Times. URL https://www.nytimes.com/2026/03/06/opinion/ ezra-klein-podcast-dean-ball.html. Accessed March 8, 2026.

Kokotajlo, Daniel, Scott Alexander, Thomas Larsen, Eli Lifland and Romeo Dean (April 2025). “AI 2027.” Webpage. URL https://ai-2027.com/.

Land, Nick (2011). Fanged Noumena: Collected Writings 1987–2007. Cambridge, Massachusetts: MIT Press.

LeCun, Yann, Yoshua Bengio and Geoffrey Hinton (2015). “Deep Learning.” Nature, 521: 436–444. doi:10.1038/nature14539.

Lin, Zhen, Shubhendu Trivedi and Jimeng Sun (2024). "Generating with Confidence: Uncertainty Quantification for Black-box Large Language Models", Transactions on Machine Learning Research, https://arxiv.org/abs/2305.19187

Lindblom, Charles E. (1965). The Intelligence of Democracy: Decision Making through Mutual Adjustment. New York: Free Press.

Liu, Xiaoou, Tiejin Chen, Longchao Da, Chacha Chen, Zhen Lin and Hua Wei (2025). "Uncertainty Quantification and Confidence Calibration in Large Language Models: A Survey", pp. 6107--6117 in L. Antonie et al. (eds.), KDD '25: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining, doi:10.1145/3711896.3736569, https://arxiv.org/abs/2503.15850

Lovecraft, H. P. (2005). “At the Mountains of Madness.” In Tales (Peter Straub, ed.), pp. 481–586. New York: Library of America.

Lu, Donna (Jan. 2025). “We Tried Out DeepSeek. It Worked Well, Until We Asked It About Tiananmen Square and Taiwan.” The Guardian. URL https://www.theguardian.com/technology/2025/jan/28/we-tried-out-deepseek-it-works-well-until-we-asked-it-about-tiananmen-square-and-taiwan. Accessed: 2026-02-08.

Manson, Katrina (2026), “Google to Provide Pentagon With AI Agents for Unclassified Work,” Bloomberg, March 10, 2026.

Matias, J. Nathan (2023). “Humans and Algorithms Work Together. So Study Them Together,” Nature, 617, 248-251. doi: https://doi.org/10.1038/d41586-023-01521-z

Mohri, Christopher and Tatsunori Hashimoto (2024). "Language Models with Conformal Factuality Guarantees", pp. 36029--36047 in R. Salakhutdinov et al. (eds.), Proceedings of the 41st International Conference on Machine Learning [ICML 2024], https://arxiv.org/abs/2402.10978

McAuley, Paul (2010). Into Everywhere. London: Orion.

McCulloch, Warren S. (1945). “A Heterarchy of Values Determined by the Topology of Nervous Nets.” Bulletin of Mathematical Biophysics, 7: 89–93. doi:10.1007/BF02478457. Reprinted in (McCulloch, 1965, pp. 40–45).

 — (1965). Embodiments of Mind. Cambridge, Massachusetts: MIT Press.

Naidu, Suresh (Feb 2020). “In Praise of Blindspots.” LPE Project Blog. URL https://lpeproject.org/blog/in-praise-of-blindspots/. Accessed: March 8, 2026.

Narayanan, Arvind (2026). “What If Algorithmic Fairness Is a Category Error?” In Contemporary Debates in the Ethics of Artificial Intelligence (Sven Nyholm and Atoosa Kasirzadeh and John Zerilli, eds.), pp. 77–96. Wiley-Blackwell. URL https://www.cs.princeton.edu/~arvindn/ publications/algorithmic_fairness_category_error.pdf. Accessed: 2026-02-08.

Nelson, Alondra (forthcoming). “Field Theory: AI as Social Science Question, Object, & Tool.” Daedalus.

Newell, Allen and Herbert A. Simon (1956). Current Developments in Complex Information Processing. Tech. Rep. P-850, Rand Corporation. URL http://doi.library.cmu.edu/10.1184/pmc/simon/box00006/ fld00396/bdl0001/doc0001.

Nguyen, C. Thi (2024). “The limits of data.” Issues in Science and Technology, 40(2): 94–101. URL https://www.jstor.org/stable/27395455.

Origgi, Gloria (2017). Reputation: What It Is and Why It Matters. Princeton, New Jersey: Princeton University Press. doi:10.1515/9781400888597. Translated by Stephen Holmes and Noga Arikha from La réputation, qui dit quoi de qui (Paris: Presses Universitaires de France, 2015).

Perrin, Andrew J. and Katherine McFarland (2011). “Social Theory and Public Opinion.” Annual Review of Sociology, 37: 87–107. doi:10.1146/annurev.soc.012809.102659.

Quach, Victor, Adam Fisch, Tal Schuster, Adam Yala, Jae Ho Sohn, Tommi S. Jaakkola and Regina Barzilay (2024). "Conformal Language Modeling”, pp. 11654--11681 in B. Kim et al. (eds.), International Conference on Learning Representations 2024 [ICLR 2024], https://arxiv.org/abs/2306.10193

Raginsky, Maxim (December 2025). “The Paradox of Highly Optimized Tolerance.” Weblog post, The Art of the Realizable. URL https://realizable. substack.com/p/the-paradox-of-highly-optimized-tolerance.

Rahimi, Ali and Ben Recht (December 2017). “Reflections on Random Kitchen Sinks.” Weblog post, argmin.net. URL https://archives.argmin.net/2017/12/05/kitchen-sinks/. Accessed 05-08-2026. 

Recht, Ben (Aug. 2023). “Are there always trade offs?” arg min. URL https://www.argmin.net/p/are-there-always-trade-offs. Accessed: 2026-02-08.

Reynolds, Alastair (2000). Revelation Space. London: Gollancz.

Scott, James C. (1985). Weapons of the Weak: Everyday Forms of Peasant Resistance. New Haven, Connecticut: Yale University Press. doi:10.12987/9780300153620.

 — (1998). Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed. New Haven, Connecticut: Yale University Press. doi:10.12987/9780300128789.

Shalizi, Cosma (November 2010). “The Singularity in Our Past Light-Cone.” Weblog post, Three-Toed Sloth. URL https://bactra.org/weblog/699. html.

Simon, Herbert A. (1957). Models of Man. New York: John Wiley.

 — (1996). The Sciences of the Artificial. Cambridge, Massachusetts: MIT Press, 3rd edn. doi:10.7551/mitpress/12107.001.0001. First edition 1969.

 — (1997). Administrative Behavior: A Study of Decision-Making Processes in Administrative Organizations. New York: Free Press, 4th edn.

Simon, Herbert A, Donald W. Smithburg and Victor A. Thompson. (1958). Public Administration. New York: Alfred A. Knopf.

Spufford, Francis (2017). “Idols of the Marketplace.” In True Stories, pp. 110–119. New Haven, Connecticut: Yale University Press. doi:10.12987/9780300231601.

Stross, Charles (2005). Accelerando. London: Orbit.

Therborn, Göran (1978). What Does the Ruling Class Do When It Rules?. London: New Left Books.

Thompson, E.P. (1963). The Making of the English Working Class. London: Gollancz.

Tilly, Charles (1984). Big Structures, Large Processes, Huge Comparisons. New York: Russell Sage Foundation.

Valente, Catherynne M. (Oct. 2011). “Silently and Very Fast.” Clarkesworld Magazine, (61). URL https://clarkesworldmagazine.com/valente_10_ 11/. Accessed 03-08-2026.

Vaswani, Ashish, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin (2017). “Attention Is All You Need.” In Advances in Neural Information Processing Systems 30 [NIPS 2017] (I. Guyon and U. Von Luxburg and S. Bengio and H. Wallach and R. Fergus and S. Vishwanathan and R. Garnett, eds.), pp. 5998–6008. Red Hook, New York: Curran Associates. URL http://arxiv.org/abs/ 1706.03762.

Vinge, Vernor (1992). A Fire Upon the Deep. New York: Tor.

 — (1993). “The Coming Technological Singularity: How to Survive in the Post-Human Era.” In Vision 21: Interdisciplinary Science and Engineering in the Era of Cyberspace, no. CP-10129 in NASA Conference Publication, pp. 11–22. NASA Lewis Research Center. URL https://edoras.sdsu.edu/~vinge/ misc/singularity.html. Accessed March 8, 2025.

Vovk, Vladimir, Alexander Gammerman and Glenn Shafer (2005). Algorithmic Learning in a Random World (New York: Springer-Verlag), doi:10.1007/b106715

Wallace, Jeremy L. (2014). “Juking the Stats: Authoritarian Information Problems in China.” British Journal of Political Science, 46: 11–29. doi:10.1017/S0007123414000106.

Yang, Eddie (2026). “The Limits of AI for Authoritarian Control.” American Journal of Political Science, forthcoming. URL https://www.eddieyang. net/research/AI_limits.pdf. doi:10.1111/ajps.70045.

Yates, JoAnne (1993). Control Through Communication. The Rise of System in American Management. Baltimore: Johns Hopkins University Press.

Yiu, Eunice, Eliza Kosoy and Alison Gopnik (2024). “Transmission Versus Truth, Imitation Versus Innovation: What Children Can Do That Large Language and Language-and-Vision Models Cannot (Yet).” Perspectives on Psychological Science, 19(5), 874-883. doi: 10.1177/17456916231201401

Acknowledgments

Earlier versions of this paper were presented at the 2025 meeting of the Società Italiana di Scienza Politica, and the 2026 workshop, “Cultural AI: An Emerging Field,” at the Remarque Institute at New York University, and at the University of Florida. We are grateful both to the participants at these events, and to Melanie Mitchell, Cris Moore and Jack Shanahan for comments and feedback.

© 2026, Henry Farrell and Cosma Rohilla Shalizi

Cite as: Henry Farrell and Cosma Rohilla Shalizi, AI as Social Technology, 26-5 Knight First Amend. Inst. (May 11, 2026), https://knightcolumbia.org/content/ai-as-social-technology [https://perma.cc/25BX-GUQL]. 

“There’s increasing awareness of the crucial role lawyers play during periods of democratic backsliding, in the United States and elsewhere, but not nearly enough public examination of it,” said Katy Glenn Bass, research director at the Knight First Amendment Institute. “We created this podcast to open that space, surface questions for further scholarship, and trace what responsibility looks like for lawyers in this moment.”

Authoritarianism is often understood as lawless, marked by constitutional rupture or institutional breakdown. But many of the most effective assaults on democracy operate through legal systems themselves. Around the world, leaders have used laws, courts, and institutional processes to consolidate power, restrict dissent, and weaken accountability while maintaining the appearance of legality. Similar pressures are emerging in the United States, raising urgent questions about how the legal profession responds when these dynamics take hold closer to home.

“One of the central concerns that motivated this project is that we’ve spent a great deal of time studying courts and political institutions, but far less time examining lawyers themselves,” said Madhav Khosla, senior fellow at the Knight First Amendment Institute and professor at Columbia Law School. “Lawyers are often the first to see how legal systems are being used or misused, yet we haven’t fully reckoned with the role they play in shaping those outcomes.”

The podcast is part of a broader collaboration between the Knight Institute and Khosla examining how the legal profession is evolving in response to democratic backsliding. The project includes a scholarly essay series and public convenings exploring how professional norms, ethical frameworks, and institutions such as bar associations respond to political pressure.

The first episode, “What Does Legal Authoritarianism Look Like?,” features Kim Lane Scheppele, the Laurance S. Rockefeller Professor of Sociology and International Affairs at the Princeton School of Public and International Affairs. The episode explores how authoritarianism can take hold through ordinary legal processes and what it means for lawyers when legal systems depart from their constitutional commitments.

Future episodes will feature leading scholars and practitioners, including Lawrence Lessig, Deborah Pearlstein, and David Dyzenhaus examining the role of the legal profession across a range of political and institutional contexts.

“Conversations about democratic backsliding and the role of law can feel abstract, even as these pressures are playing out in real time,” said Candace White, deputy director of communications and digital strategy at the Knight First Amendment Institute and executive producer of “Lawyering Without Law.” “This series makes those pressures visible, helping people understand how legal systems are being used and what’s at stake.”

“Lawyering Without Law” is available on Apple Podcasts, Spotify, and other platforms, with new episodes released every other Friday. The podcast is hosted by Katy Glenn Bass and Madhav Khosla and produced and engineered by Dustin Foote, with Candace White serving as executive producer. Fact-checking is by Connor Menzies and Sophia Rojas; music is from Envato Elements, and artwork is by Jay Vollmar.

Listen to Episode 1 here.

Learn more about the series here.

Read more about Khosla’s research project here.

The Knight Institute has also produced four award-winning podcasts: “What Happens When Social Media Collides with the First Amendment?,” “War & Speech,” “Speech & the Border,” and “The Bully’s Pulpit: Trump v. the First Amendment.” Explore them here.

For more information, contact: Adriana Lamirande, adriana.lamirande@knightcolumbia.org 

Hi, everyone. A quick note before we start. This conversation was recorded before Viktor Orbán lost his reelection bid for Prime Minister of Hungary. After 16 years in power, he was defeated by Péter Magyar. Okay, now let’s start the show.

Kim Lane Scheppele:

The dominant way that democracies fail is through aspirational autocrats being elected in free and fair elections, coming to power, changing the law, usually in illegal fashion, until the next thing you know, there’s no way to get rid of the autocrat.

Katy Glenn Bass:

Hello and welcome to Lawyering Without Law, a new podcast by the Knight First Amendment Institute where we’ll explore the unique and important role that lawyers and the legal profession play in defending democracy or facilitating a country slide into authoritarianism. I’m Katy Glenn Bass, the research director at the Knight First Amendment Institute, and you’ve just heard a clip of a conversation with our first guest, Kim Lane Scheppele, who we’ll introduce in just a minute.

Joining me in co-hosting this podcast is Professor Madhav Khosla, the B. R. Ambedkar Professor of Indian Constitutional Law and Professor of Political Science at Columbia University and the Knight Institute’s senior fellow. Madhav, I’m really looking forward to hosting these conversations with you.

Madhav Khosla:

Thank you, Katy. I’m really looking forward to it as well. Over the next six episodes, we’ll bring both historical examples, comparisons and contrast to what we’re seeing now, as well as more specific conversations about democratic backsliding, both in the US and around the world, and in particular, the legal profession’s role in these developments over the last few years.

Joining us today on our very first episode to explore the use and deployment of the law and legal systems as a central feature of contemporary authoritarian populism is Kim Lane Scheppele, the Laurance S. Rockefeller Professor of Sociology and International Affairs at Princeton University. Professor Scheppele’s primary field is the sociology of law, and a lot of her research examines the rise and fall of constitutional governments.

She’s worked extensively on the undermining of civil liberties during the post 9/11 period, and in recent times has been integral to debates on the transition from democracy to authoritarianism. Professor Scheppele, welcome to Lawyering Without Law, and thank you for joining us.

Kim Lane Scheppele:

It’s wonderful to be here, although the thought of being without law is vaguely terrifying.

Madhav Khosla:

It is, and hopefully we can figure out exactly how that happens. So just to start us off, Kim, I was wondering if you could just share a little bit about your contemporary work on illiberal government and on democratic backsliding.

Kim Lane Scheppele:

Well, when I started working on this line of research, I actually didn’t think it was going to be about backsliding democracies. I thought I was looking at, well, I don’t know, frontsliding democracies, countries that were coming out from under Soviet tutelage that were setting up democracies first in what was Eastern Europe, then we called it Central Europe, and then in the former Soviet Union itself.

I worked for four years at the Hungarian Constitutional Court in the 1990s. I worked for one year at the Russian Constitutional Court in the early 2000s, and I thought what I was watching was the onset of democratic institutionalism, rule of law, and respect for rights. Now, oh, I guess I shouldn’t have been massively surprised. Both Viktor Orbán and Vladimir Putin, once they came to power, were lawyers and they understood how law worked.

And so as a result, when they decided they wanted to stay in power for the long term and not allow power to rotate, they wound up dismantling the very institutions that had just been set up and they did so by law. Because there were lawyers, they could work out what are the soft spots in the system, that if you pushed on them would generate dictatorship rather than democracy.

They worked with their parliaments to push laws through. They worked with courts to increase the number of judges actually, often on courts, packing them with their own people, and using perfectly constitutional measures or what looked to the outside like perfectly constitutional measures. They were able to little by little lockdown dictatorship in their hands.

But if all you looked at was the form, you wouldn’t have realized that dictatorship was coming back in those places. You would’ve thought, “Ah, a new leader is elected and he’s changing the law in order to fulfill his mandate.” So it took quite a lot for other people to catch on. But that leading edge, shall we say, of dictatorship by law turned out to be a transportable model.

You could pack it up and you could move it to other countries. And now that’s the dominant way that democracies fail is through aspirational autocrats being elected in free and fair elections, coming to power, changing the law, usually in illegal fashion, until the next thing you know, there’s no way to get rid of the autocrat.

Katy Glenn Bass:

I know you’ve written about this a lot, this concept of autocratic legalism. And is it fair to say that when you first started describing that or sort of following that story, it was something that was characteristic of new democracies or countries transitioning out of authoritarianism?

Kim Lane Scheppele:

Yeah, those were the ones I focused on first, but then it turned out that it spreads.

Katy Glenn Bass:

Right.

Kim Lane Scheppele:

And in fact, at the beginning, I think a lot of analysts didn’t take this movement all that seriously precisely because it was easy for them to say, “These are new democracies.” They weren’t really established. Easy to knock over. And so therefore, what happens in new democracy stays in new democracies, and longstanding democracies are relatively immune from all that, but now we know that that’s just not true.

Katy Glenn Bass:

So this concept of autocratic legalism and this belief that it was something confined to new democracies, how do you see this now that we’re in 2026? Is this what the United States and other established democracies are beginning to grapple with, this same sense of autocrats manipulating the law?

Kim Lane Scheppele:

Yeah. So I think we’re certainly seeing that in the United States. What we see is that President Trump elected in a free and fair election comes to power and he issued I believe it was something like 325 executive orders in his first year. The only president who had issued more than that was Franklin Roosevelt who was also involved in a constitutional revolution. These same methods can be used whether you like the revolution or not.

And now what we see is those executive orders being litigated up to a friendly Supreme Court that Trump had an opportunity to, shall we say, change the composition of in his first term. And if all of that succeeds, if the Supreme Court goes along with the executive orders over existing statutes and over prevailing interpretations of the Constitution, we’re going to find ourselves with a new Constitution.

Now, let me just say this isn’t something Trump dreamed up on his own. First of all, I think we attribute way too much to him when he’s out playing golf most of the time. This is something that really was a kind of blueprint that came from The Heritage Foundation Project 2025. You can see him just marching along through all of the things Project 2025 identified. But it turns out that Project 2025 was not just an American project.

As it turns out, Viktor Orbán, Prime Minister of Hungary, and his big think tank, the Mathias Corvinus Collegium is the big mothership in Hungary, and the similarities are so strong between what Orbán did to take over his government in 2010 and what Trump is doing in his first year. That when I read Project 2025, the first thing I said was, “Wow, somebody’s really studied Hungary closely.” And sure enough, what we’ve seen is the very same script being applied here.

To fill that in, let me tell you just two things that were crucial in Orbán’s first year. The first thing he did was to come into power. And while he’d been out of power, he analyzed the national budget to figure out what are all the ways this budget supports people who are likely to oppose me. Comes into power and immediately slashes the money for everything. Puts the opposition into disarray. He defunds them.

They’re running around trying to get their money back and not paying attention to what comes next, which is this shoveling through tons and tons of laws. So what happens again, the other thing that Orbán did was to come in, suspend the civil service law, mass fire public employees so that he could then reinstate the civil service law, hire his own cronies into those positions, and have an administrative state that was not going to stand up to him.

So just those two things alone account for a big chunk of the disorientation when Trump came in those first few months. So really the same kind of playbook happening again right here in the US.

Madhav Khosla:

So Kim, one of the things that I think somewhat follows from your important insight is that we are seeing changes through law, but we are also therefore in some ways seeing changes through lawyers. And lawyers in a way are central to the changes that are happening.

And even before these developments became front and center in our lives, you’ve actually spent a lot of time thinking about how lawyers may not actually be the natural allies of liberal democracy. And some of your work has challenged the conventional understanding of what academics refer to as the legal complex and things like that. And could you walk us just a little bit through this?

Kim Lane Scheppele:

Yeah. So first of all, Orbán and Putin themselves were lawyers. A lot of these autocrats that come to power either are lawyers or they have lawyers that they’re working with in close association. And part of that is because, as I put it, I have a chapter title in this book that I’m almost done with, the book, by the way, is called Destroying Democracy By Law, right on point to your podcast.

One of the chapter titles is called Law is the way the State talks to itself. Now, the reason why I use that formulation is that it’s one thing as king to command somebody to do something in particular, but the only way you can command large numbers of people effectively is to give them general orders. And those general orders are what law actually is, particularly administrative law, particularly orders to the administrative state itself.

But also by entrenching ideas in law, it means that you get followers of two kinds. One of the kinds that were already on board with your political program, they do anything you told them to anyway. But the second thing is you get on board the rule followers, the people who say, “Because something is law, therefore we have to obey it.” And any successful democracy should have a large number of people in that second category.

That’s how this whole thing works. And so if you change the law, people may say, “Well, I disagree with it, but it was made by an elected official in a proper process and therefore I am bound by it.” It’s exactly what you want citizens to think. These are good citizens. So when you put into that process laws that actually have the effect of undermining democratic institutions, you get a lot more buy-in than if you simply did it by edict or by force.

And so it legitimates the entire enterprise of undermining democratic institutions. So that’s why law becomes so crucial. Now, who are the lawyers who do this stuff? Some of the lawyers who do this are lawyers who believe in the leader and believe in the program. But oftentimes you get lawyers doing this because law is like a giant math problem if you like math.

It’s just fun to play with the rules and think, well, gee, how would you do this? How would you do that? Is there a way you can make an argument for this, that, and the other thing? And frankly, legal education in many places, and here I include the US, lends itself toward exactly this kind of lawyering, because how often do we take first year law students and say, “Argue the opposite of what you believe? Come up with the best argument for the position that you don’t have."

How often do we say that lawyers in private practice have the highest duty to represent their clients within the boundaries of law, but often making creative legal arguments that would lead to a result that the lawyers themselves would not choose? So yes, we have legal ethics. Yes, lawyers have some boundaries, but think about what legal ethics looks like.

Very few of the rules are about the values that lawyers are supposed to uphold. So yes, they’re supposed to uphold the Constitution, that’s the main values-based rule, but what if the Constitution is changing? What if the new political movement says the Constitution’s something else? You can swear allegiance to a changing constitution and be completely on board with an autocratic project.

So lawyers have learned to be hired guns. They’ve learned to think that clever legal arguments are good lawyering, and they have much less training and much less grounding in the values that underwrite Democratic institutions so that the lawyers themselves may not realize how much they’re undermining Democratic institutions with the very things they’re recommending.

Katy Glenn Bass:

So I think that’s exactly right. And we’ve been thinking a lot about the role of legal education in terms of what young law students are told when they’re in training in terms of what the expectations are and the duty to zealously represent clients. And I’m wondering, are there comparative examples of different ways of educating young lawyers that you know of from other countries or what would you do if you were in charge of reforming legal education in the United States to try to remedy some of those problems, especially in such a polarized time?

Kim Lane Scheppele:

Right. So this is very difficult because in the United States, what we teach is, yes, you are bound to the Constitution. You swear an oath to the Constitution, not to the president, and yet the Constitution itself, how do we know what the Constitution is? And the answer has been, the Supreme Court tells us that. But what if the Supreme Court’s been captured, especially by a faction that isn’t committed to democracy?

The Supreme Court starts issuing cases with which you disagree, but you say, “My oath to the Constitution is an oath to the Supreme Court.” We don’t have a way for lawyers to distinguish between a Democratic constitutional vision and contingent decisions of a court that may have come under political direction. And so this is what you see happening right now in the US.

I have two sets of lawyer friends. One is the set of lawyer friends, and mostly these are people who have international experience because they’ve seen this happen elsewhere. They say, “Wait a second, this is a counter-constitutional project. The point of this is to unsettle and undo the constitutional orders that we have. So therefore, I’m going to go into the resistance and fight to prevent that from happening."

So that’s a small set of lawyers. The other lawyers are saying, “Well, we’re still doing law. We’re still litigating. We’re still pretending like all this is real. Yes, we’re disappointed with Supreme Court decisions, but we’ve learned to live with the ones and then you go back to fight another day.” And everybody is assuming that all the judges up and down the system are still doing law. And that’s the big question.

What happens when you get autocratic capture? One of the first institutions the autocrats go after are the courts precisely to take advantage of the fact that lawyers are trained to follow them to legal education. So then how do we train lawyers elsewise? This is where it seems to me we have to be thinking about courses in democratic and constitutional theory that are not tied to court decisions, that are all about the point of the enterprise.

So let me just tell you one course I used to teach when I was a law professor. I was a law professor at Penn for a decade. We were trying to reform legal education as every law school keeps doing. We decided that students were reading too many cases in their first year, that they needed a break from that. We had an interdisciplinary faculty, so we put in place this requirement that students had to take one elective that would get their nose out of case law.

So I decided to offer a course, which was a role play of Madison’s notes on the Constitutional Convention. And if you read that, it’s a very dense... It’s like reading the phone book, frankly. You have to really push yourself through it because it’s just a lot of keeping track of who said what and exactly what the votes were, and it’s really hard. Turns out if you do it as a role play, take it as a script for a play and you assign students to play all the different parts.

So somebody’s Elbridge Gerry and somebody’s Ben Franklin and somebody’s James Madison, et cetera. And then what happens is the book comes alive. So what I would do with my students, I’d get them to take up these various roles. Then I’d have them go off and read biographies of their characters, read something about their states, how they were elected to serve in the Constitutional Convention.

And then each person had to keep track of what their character said through the debates and we kind of reenacted them in class. And then when we got through the whole thing, these delegates had to go back to their states and work their way through the ratification conventions and report back to the class on what happened in the ratification conventions.

So I thought of it as my anti-originalism course, because once you’ve done that, you realize there were so many different views on offer, so many different debates, that the idea that the framers were one kind of thing just is belied by a simple reading of all of this. But I realized in retrospect that part of what we were doing in that class was really asking the question, what should constitutions do?

And I think it empowered students both to challenge even the Supreme Court’s version of originalism now, but to recognize that in any robust constitutional democracy, you’re going to have a lot of different views on offer and some of them lead down the path to dictatorship and some of them allow you to maintain your democracy. So that was just one example of a course that you could teach that doesn’t put the Supreme Court in the middle of what the Constitution means, gives people resources to fight back.

Madhav Khosla:

One of the other things that I think that is sort of worth perhaps talking about is that the legal profession is self-regulated like the professions are in modern societies. And how effective do you think that self-regulation has actually been in curbing the problem in the countries that you’ve studied?

Because you’re absolutely right that we don’t really have a system where lawyers are taught to distinguish between ordinary democratic moves and a kind of democratic backsliding. I was just wondering if, for instance, in Hungary, has the bar pushed back at all against Orbán or the lawyers who’ve supported his efforts? Or for example, have you seen actually the bar refusing to represent people who might be against the government?

Kim Lane Scheppele:

Well, unfortunately, no, not in Hungary, not in Russia. They all caved. And part of this is because if what you’re doing as a lawyer is winning cases in court and the courts come to be captured, fighting back against the system just guarantees that your clients are going to lose, right? So you need the bar to have a bigger field of vision than just representing their clients.

Occasionally what’ll happen is you’ll get in some places the bar pushing back. So for example, in Poland, the bar pushed back. This was a pro-autocratic government came in in 2015. They immediately started attacking the judiciary. One of the things that they were attacking was, among other things, the role of the bar in picking judges, which they diminished. But it was also the case that they were using unlawful methods to put judges onto the constitutional tribunal.

The bar actually stood up to that, as did a number of judges. And so what you’d see are these lawyers marches where everybody would show up in their legal robes, the kind of things they’d appear in court in, and they would march in the streets with signs saying, “Save our judges.” It was really quite moving. And when the president of the Supreme Court was fired, she insisted that she’d been fired unlawfully.

She decided to go to work as if she had not been fired. Hundreds of lawyers showed up to escort her into the courthouse. So there are moments when you see the bar really rising to the occasion.

Malcolm Brabant:

Free courts now is the Clarion cry. Outside a courthouse in Central Warsaw, demonstrators demand the removal of a judge appointed by the populist conservative government to replace one of a more independent spirit. They accuse the country’s justice minister of being a judicial puppet master.

Michal Wawrykiewicz:

We are still on the battle for the rule of law in Poland. The rule of law is...

Kim Lane Scheppele:

One thing that I think helped that in Poland was the fact that Poland is part of the EU. There are now sort of EU wide bar associations, so that if your country is coming under attack, the next country over may not be, and you get sort of an echo back from people outside your system about what your system looks like to others. And so there were a number of these bar groups that then went to court to challenge the fact that the European Union was not doing enough to preserve judicial independence in Poland.

So again, it was lawyers associations across Europe that did that. Now, it helped that Poland was the second case. Hungary was the first. When the Hungarian judiciary was being compromised, it was so novel people didn’t know what they were seeing. But just to bring this back to the US, one of the interesting things that happened here well before Trump came back the second time, I’ve been going around talking to different bar associations because they all think all these comparative examples are kind of interesting.

And one thing I started doing in some of these bar association talks was to go in and look at what big law did in those countries when the autocratic consolidation started. And the answer was that they turned tail and ran. They often turned over their legal business to local lawyers and they just shut down their offices. So here’s the thing, big law was not brave in other countries, these big international law firms with thousands of lawyers with offices in every major capital.

When the autocratic impulse came over these rulers, the big law firms hung in there for a while, and then they left. What do we expect them to do in the US if that’s what they did abroad? And I’m a little surprised because I would talk to these bar groups and literally name the firms that left Hungary, that left Poland, left Russia, et cetera, and no response. It was so interesting.

So from that, I guess I wasn’t surprised to see what’s happened here, which is yes, Trump singled out law firms with executive orders and some of them caved and some of them fought. That’s usually the story. But what really happened was that all the law firms, including the ones that fought, have really avoided getting into the trenches, fighting against Trump’s executive orders.

They’ll still defend their clients even if their clients are leaned on by Trump, but they won’t put their necks out and actually start to figure out how to use law creatively to push back against the constitutional revolution that we’re seeing. Big law is just all out.

And the bar associations, when they get complaints that, for example, that there are DOJ prosecutors who have violated basic legal ethics, you can actually list the numbers of the rules that they violated, the bar associations just say, “Oh, we’re going to wait for a criminal procedure before we look at what our lawyers have done,” or they start an endless investigation that goes down a rabbit hole and never comes out.

So there’s a couple of bar associations that have disbarred people who have obviously acted in violation of their oaths, but not very many. So the US bar is, shall we say, protecting itself and not protecting the Constitution.

Katy Glenn Bass:

That’s really interesting. I didn’t know that story of the big law firms in the other countries. Can I ask, I mean, how different or not is the situation of US civil society where the bar in the US, lawyers like Knight Institute lawyers where we are bringing cases from a more independent neither government nor a big law stance, is there a difference in terms of the size and the relative wealth of US civil society versus some of these other countries? And do you think that has had any meaningful difference?

Kim Lane Scheppele:

Yeah. So one of the things about established democracies as opposed to new ones is that the established ones tend to have bigger and deeper civil society organizations. And so what you see in the US is Knight Institute, Democracy Defenders, Democracy Forward, Public Citizen, Protect Democracy, ACLU, and Democracy Docket, Marc Elias, from doing all the voting stuff, what you see is there’s a good half dozen to dozen groups like that that are carrying the heavy load of fighting back in court because the lower court judges are still holding the line even if the Supreme Court’s not.

And it’s worth litigating, it’s worth fighting this stuff just to throw sand in the gears, slow things down. We forget that there’s about 150, 160 injunctions that are still holding out there. A recent report showed that of all those big science grants that were cut at the beginning of 2025, by the end of the year, 80% of them have been restored through litigation. So litigation is still an important path.

And it’s these NGO interest group, mid-sized, overworked, stressed out groups of lawyers that are handling the huge fight. I’m struck by the big difference between what’s happening now and what happened after 9/11. So after 9/11, there was a lot of overreaction on the part of the Bush administration. There was some civil society pushback, but then the torture memo came out.

Then it was apparent that the US government was aiding and abetting a program of what they were calling enhanced interrogation. There were renditions and detentions that were being done in violation of basic due process norms of international law. And so as soon as that came out, it turns out that big law was all in. Every single firm, from left to right, devoted pro bono work to torture victims, to legal detentions, to those rendition cases.

I know because I actually was running a LISTSERV for all the lawyers that were involved in all that work to pushing back against the Bush administration and the war on terror. It could not be more different now. Big law has just gone silent. What you see now are partners that are quitting. Some of them are forming their own firms, but they’re small firms. They don’t have the immense deep pockets of big law.

So I feel like from a legal standpoint, we’re fighting the onset of dictatorship in the United States with one hand tied behind our back by lawyers who insist upon thinking that we’re still in some kind of normal politics. And I’m very worried about that. If the bar isn’t there to defend the Constitution, the legal order, basic rights, due process, then what’s the average person to do who doesn’t have the skillset to even know how to name those things?

So yes, I’m a little disappointed in the organized bar. We need to do a whole lot better than that if we’re going to save democracy and lawyers are going to be crucial to that, even if they can’t do it alone.

Madhav Khosla:

Kim, do you think some of that potentially has to do with the fact that a lot of people in big law don’t see themselves as lawyers in the way that you and I conceive of lawyers? Maybe actually like the people who you mentioned in Hungary, maybe they just see themselves as basically figures engaged in transactions rather than your Blackstonian advocate who’s meant to stand up for things.

I mean, maybe it’s not fortuitous that the firms that actually stood up against Trump are those with large litigation practices. And so actually a lot of what you and I think of as big law is basically just bankers in disguise or something like that. I mean, I’m obviously putting the point provocatively, but I was wondering whether there’s something about the changing character of the profession itself that captures some of this dynamic.

Kim Lane Scheppele:

Yeah. What you say is what actually quite a few people have been arguing, that basically lawyers are doing much more than law and they’re doing transactional work. They’re arranging their client’s finances. They’re engaged in preventative action to keep them from ever being sued. They’re organizing their paperwork essentially. So yes, that happens. But here’s what’s really interesting, it’s until gravity disappears, you don’t notice how much you rely on it.

And transactional work relies on there being a fundamentally rule of law based system that underwrites all those transactions. There’s a wonderful book by a legal anthropologist, Annelise Riles, and she studied the Japanese Stock Exchange. And what she studied, and if you think about it, unless you’re in this line of work, you never think about this problem. If you buy and sell a stock, suppose you buy a stock, somebody had to sell it.

But when you buy and sell in an exchange, you have no idea who the seller was when you buy stock. Somewhere underneath all that, there’s a whole bunch of lawyers writing contracts. And Annelise studied the backroom of the Japanese Stock Exchange and all the people, all the lawyers who were matching the buys and the sells to figure out how to do the legal contract that underwrote the exchange.

Now, you don’t see those legal contracts unless there’s some problem, but they have to be there so that there won’t be a problem. And a lot of the legal system operates invisibly like that, where it just is underwriting everything we count on. When I teach sociology of law, I teach what is marriage, what is a passport, what is money, what is Princeton University? They’re all legal forms.

They’re not something that exists apart from the legal guarantees that underwrite them. So you could be a transactional lawyer thinking that all you’re doing is just rearranging the subsidiary over here and moving assets over there and setting up some kind of firewall so that a suit won’t deprive the firm of all of its resources over here. All that stuff. But all of that stuff may not take you into a courtroom, may not take you near the government.

But if you don’t have the infrastructure of courtrooms and governments and basically the guarantee of the rule of law that a democratic government provides, all those transactions are simply paper. And I just wish lawyers realized they rely on think of it as legal gravity that holds everything down and keeps it from floating off into the air. They can’t escape that and they have a responsibility for protecting that.

Katy Glenn Bass:

Do you have any sense that lawyers in the US are becoming more aware of that legal gravity or sort of becoming more aware of that fundamental underpinning of the rule of law that they need? Is there a tipping point that we should expect?

Kim Lane Scheppele:

Well, yeah, I think we’re seeing it a little bit. I mean, so first of all, when Trump first came in, things happened so fast and people couldn’t believe what they were seeing, that they preferred not to see it. And I think now the destruction, the rubble, the Justice Department lies in rubble. Trump’s picture appears on the Justice Department. I mean, there’s so many signs that our legal order is not in good order.

There’s so many signs that we’re getting from the Supreme Court that the Supreme Court doesn’t know what the law was, or rather it doesn’t care what the law was. It will do something else. I think if you’re paying attention, what you realize is we’re very far into autocratic capture of our legal system and we’re very far into the destruction of the very institutions that will allow peaceful transfers of power in the future. Now, if people realize that...

I mean, one of the problems is that by the time you realize it, it may be too late to fix it, right? So you’re then in this situation where all you can do is patch the rubble back together again somehow. So I think we’re already so far along in the destruction of democratic institutions, but we’re now at the point where you have to realize we’re not just going to pick up the rubble of the Justice Department and try to patch it back the way it was.

We’re going to have to reimagine the kind of thing the Justice Department is. We’re not going to just be able to go to the Supreme Court and say, “Cut it out. Follow your own precedent.” We’re going to have to reimagine what a Supreme Court will look like. I mean, this is really major reconstruction, I think. So just for example, what most countries do that have been through dictatorships, one of the first things they do is they separate the Justice Ministry from the prosecution service, and they wall off the prosecution service from political influence.

Wouldn’t that be a good idea for us? We just happen to have this Justice Department because we were the first comer and all those things are packaged into one institution, but most governments separate them. There’s also another thing which is that most governments don’t have judges that stay on the bench for 30 years at their highest court level. So they have term limits.

They have age limits. They have processes that guarantee that the judges who get appointed, especially to the highest court, are named in procedures that require cross-party agreement. You could start imagining importing all these ideas that other countries have come up with. I mean, America hasn’t been great at borrowing back because we think we invented it.

But it’s like we’re driving around in a Model T Ford and everybody else has got these fancy cars that can do things that Model Ts couldn’t do. And we’re saying, “But isn’t the Model T wonderful because it came first?” Constitutional drafting has gone way past where the US is. And we’re now so far into this system of autocratic capture that one of the things I do is I teach about all these other systems.

And I left law school teaching in order to teach the undergraduates before they get to US law school. So that when they get to US law school, they’re going to say, “Really? That’s how we do things? There are all these other ways you could do things. Why do it like that?” So I’m being a subversive by getting to the undergraduates first, but I think we need to look around a little bit.

Katy Glenn Bass:

Kim, is there anything else that’s on your mind on these topics that you know so well that we haven’t touched on yet?

Kim Lane Scheppele:

Yeah, so I actually didn’t lean very much into the self-regulation question. It’s the same issue that’s now coming up with regard to academic freedom. When you have a self-regulating profession, it actually has to take self-regulation seriously. And bar associations have been sort of polite clubs. They haven’t wanted to set and enforce the rules very seriously. Yes, there are disbarment procedures.

It’s usually in egregious cases by lawyers who aren’t that powerful. You’re not seeing very powerful lawyers doing very big things getting disbarred, right? And we have the same problem with academic freedom that we’ve allowed things to happen in universities that probably shouldn’t have happened if we had been policing standards of not just disciplinary conduct, but of what universities are supposed to be for.

So we have to think about self-regulation as not just are the I’s dotted, T’s crossed in particular behavior of those who are regulated. We have to think bigger and more theoretically about the principles that we’re supposed to be defending. And there’s a reason why the bar has always been thought of as separate from ordinary jobs. It’s because it’s a normative enterprise as well as a practical one.

Americans used to think, if we swear an oath to the Constitution, that takes care of itself. What if the Constitution breaks? Then what? And we’re sort of in that moment when just following decisions of the Supreme Court as an oath of office, it may come... I should say this, this is my revolutionary statement, there may come a time when the patriotic and constitutional thing to do is to disobey a decision of the Supreme Court.

I hate to say that. Other countries have gotten to that point. You can’t just do it as an act of force and violence. You have to do it because you’re invoking constitutional principles that are more important and more enduring. And if we have no idea what those are, we don’t know what the point is at which we have to say, “Just having the Supreme Court tell us what the Constitution means is not enough."

And that’s what I wish the bar would think about. It’s a moment when lawyers should be among the bravest because the tactics that are being used to dismantle democratic institutions are legal ones. The lawyers will see what’s happening before everybody else.

We need to be the early warning signal to everybody else that this is what’s going on, and we need to play a key role in building these things back, because ultimately a government of laws is a government of laws and lawyers do have some privileged position in figuring out, designing and enacting those laws. And so where’s the bar? We need the bar.

Katy Glenn Bass:

This has been such a wonderful conversation, Kim. Thank you so much for making time to talk to us today and for sharing all of your insights from your years of scholarship and advocacy.

Kim Lane Scheppele:

Well, thank you so much for doing the whole podcast. And I really hope you get a lot of listeners.

Katy Glenn Bass:

Us too.

Madhav Khosla:

Thanks so much, Kim.

Katy Glenn Bass:

Lawyering Without Law is a production of the Knight First Amendment Institute at Columbia University and is hosted by Madhav Khosla and me, Katy Glenn Bass. This episode was produced and engineered by Dustin Foote. Fact checking by Connor Menzies and Sofia Rojas. Candace White is our executive producer. Our music comes from Envato Elements. The art for our show was designed by Jay Vollmar.

The archival clip you heard early on in the show was from PBS. Thanks to Kim Lane Scheppele who had joined us for this episode. Lawyering Without Law is available on Apple, Spotify, and wherever you listen to podcasts. Please subscribe, share, and leave a review. We’d love to know what you think. To learn more about the Knight Institute, visit our website, knightcolumbia.org.

That’s Knight with a K. And follow us on social media. We’ll see you next time for a conversation with Harvard Law Professor Lawrence Lessig. Bye for now.

I. Introduction

Transformative ideas and transformative communicative practices from around the globe have enriched American public discourse and updated how Americans interact with one another. Social media platforms are only the latest means of communicating and engaging. However, foreign-owned social media platforms that offer novel, participative ways for people to engage, communicate, and associate are a form of communications infrastructure that is not adequately protected under US law. Although there is a focus on platforms’ communicative features, there is very little discussion of their associational features.

The Supreme Court has recognized Americans’ right to access foreign ideas.1 1. Lamont v. Postmaster General, 381 U.S. 301 (1965); See also Xiangnong (George) Wang, Listeners’ Rights in the Time of Propaganda: The Story of Lamont v. Postmaster General, Knight First Amendment Institute, http://knightcolumbia.org/content/listeners-rights-in-the-time-of-propaganda-the-story-of-lamont-v-postmaster-general (last visited May 6, 2025). This right is sufficient to ensure access to mass-media forms such as print and broadcasting that depend largely on one-way transmission to audiences.2 2. Although the mass-media also encouraged engagement in the form of letters to the editor and telephoning to engage with radio and television shows, this formed a very low percentage of content consumed by the audience in comparison with social media. But it is insufficient to regulate participatory social media platforms. Together with mass media, but increasingly dominating it, social media forms the underlying infrastructure of communication in American society.3 3. See Robert C. Post, Data Privacy and Dignitary Privacy: Google Spain, the Right to Be Forgotten, and the Construction of the Public Sphere, 67 Duke L.J. 981, 1070–1 (2018) (describing Google as "communicative infrastructure"); Brett M. Frischmann, Infrastructure: The Social Value of Shared Resources 4 (describing infrastructure as "shared means to many ends") (2012). Accessing this infrastructure is distinct from, although connected to, the right to receive ideas. The right to communicative apps as infrastructure is the right to access new ways of communicating and associating. Legal discourse needs to build on the right to access foreign ideas, extending it to the right to access foreign communicative infrastructure.4 4. See Part 6 below; see also Lamont v. Postmaster General, 381 U.S. 301.

My goal in this essay is to reframe the value of foreign social media platforms (and potential future AI platforms) in the context of TikTok Inc. v. Garland as unique and creative associational and speech infrastructures and show why associational and speech infrastructures should be treated differently from other foreign infrastructure.5 5. I am responding in part to scholars who argue in favor of treating speech infrastructure like banking or energy infrastructure. See e.g., Ganesh Sitaraman, The Regulation of Foreign Platforms, 72 Stan. L. Rev. 1073 (2022). The justices adopted an outdated view of the significance of foreign social media—suggesting that a platform’s users are the only people affected by the platform—which led them to underestimate the burden on speech from restricting the app or compelling a change in its ownership. Since lawmakers and experts, including TikTok’s advocates, share this outdated view, the TikTok case is an excellent opportunity to discuss foreign platforms’ role in the American public sphere.

My argument for access to foreign communicative infrastructure extends to all significant foreign communicative apps, including chatbot-centric apps. Political, economic, and cultural influences shape communicative apps, and each app amplifies different kinds of content and offers different ways to engage. Changing foreign apps’ ownership removes these influences and burdens speech in ways that judges should take into account. Some of these questions were discussed in the context of the ‘TikTok ban’ in 2024 and 2025. Although the TikTok question has been settled, TikTok Inc. v. Garland is worth discussing because it presents the Supreme Court’s view of the role of foreign communicative apps in the American public sphere.6 6. TikTok Inc. v. Garland, 604 U. S. 56 (2025).; The Supreme Court evaluated the Protecting Americans from Foreign Adversary Controlled Applications Act in the context of an as-applied challenge brought by TikTok, and the justices clarified in their per curiam opinion that their “analysis must be understood to be narrowly focused in light of these circumstances.” Nevertheless, the case is indicative of how the Supreme Court applies the First Amendment to laws regulating foreign social media platforms. It appears that the Supreme Court will permit Congress and potentially the president to restrict Americans’ access to Chinese-made apps. Unfortunately, the judgment's implications for social media regulation have been eclipsed by the events that followed it. The justices’ application of the intermediate scrutiny test was based on a dangerously flawed understanding of social media.7 7. I am not discussing the question of whether the strict scrutiny test should have been applied since others have already done so. See e.g., Anupam Chander, Gautam Hans & Edward Lee, TikTok v. Garland Opens the Door to Global Censorship, Lawfare (2025), https://www.lawfaremedia.org/article/tiktok-v.-garland-opens-the-door-to-global-censorship. I focus on the meaning and consequences of the TikTok ruling for the American public sphere, not the actual TikTok divestiture.8 8. Executive Orders, Saving TikTok While Protecting National Security, The White House (Sept. 25, 2025), https://www.whitehouse.gov/presidential-actions/2025/09/saving-tiktok-while-protecting-national-security/.

II. Communicative Apps as Infrastructure

Infrastructure such as language, script, the press, and broadcasting make human communicative activity possible. Such infrastructure increases shared knowledge and shapes public opinion.9 9. Post, supra note 3 at 1020-1023. Online communicative platforms extend this ability by enabling Americans to share, modify, and consume content in conversation with each other and people around the world. In so doing, the platforms facilitate public discourse, including discourse on other platforms and in the media.

As “shared means to many ends,” most communicative apps are infrastructure.10 10. See Frischmann, supra note 3 at 4. Although other legal scholars have made this argument,11 11. K. Sabeel Rahman, The New Utilities: Private Power, Social Infrastructure, and the Revival of the Public Utility Concept, 39 Cardozo L. Rev. 1621 (2017); Julie E. Cohen, Between Truth and Power (2019); Julie E. Cohen, Infrastructuring the Digital Public Sphere, 25 Yale J.L. & Tech 1 (2023); Evelyn Douek & Genevieve Lakier, Lochner.Com, 138 Harv. L. Rev. 100 (2024). more work is necessary on what this means for the sectors these apps serve as infrastructure. In this essay, I discuss the implications for speech of treating significant social media platforms as communicative infrastructure. I use the term “communicative infrastructure”12 12. See Gabriel Tarde, The Public and the Crowd (1901), in Gabriel Tarde on Communication and Social Influence , 52–55 (1969); Post, supra note 3 at 1016, 1046. to refer to both the media ecosystem as a whole and its individual parts (e.g., The New York Times, the BBC, TikTok, Google Search, and Google’s AI Overview). However, my focus is on the digital public sphere, which consists of layered infrastructures that are regulated differently, including telecommunications and the open internet.

Legal scholarship on social media’s infrastructural features tends to focus on how platforms are shaped by the market, political institutions, legal frameworks, and public policy, and how they use the data they collect.13 13. See e.g., Sitaraman, supra note 5; Zephyr Teachout & Lina Khan, Market Structure and Political Law: A Taxonomy of Power, SSRN Journal (2014), http://www.ssrn.com/abstract=2490525. It facilitates a granular discussion of platforms as infrastructure, and specifically of how platforms’ infrastructural role enables public discourse by affecting which people communicate with each other and how they communicate.14 14. Post, supra note 12 at 1016, 1046 (arguing that platforms are communicative infrastructure). I aim to develop the speech-enabling aspect of online communicative platforms’ infrastructural role. Treating communicative apps as infrastructure implies recognizing that these apps have downstream effects beyond the platform.15 15. Frischmann, supra note 3 at 63–5 (discussing how the social benefit of infrastructure derives from its downstream uses). My goal is to contribute to First Amendment scholarship while integrating the valuable insights that scholars of law and political economy offer.

Communicative apps are “structured arrangement[s] that facilitate human [communicative] activity across space and time.”16 16. I have adapted Professor Julie E. Cohen’s definition of infrastructure for communicative infrastructure. See Cohen, Infrastructuring the Digital Public Sphere (2023), supra note 11 at 4. For a discussion of platforms and infrastructure, see Cohen (2019) 40-41. In addition to focusing on the influences on and design of these arrangements, and their harmful downstream effects, it is worth focusing on the extent to which they enable democratic goals.17 17. See Cohen (2023), supra note 16 at 4 (Cohen points out that data collection and targeted advertisement is among the key goals of these apps, which I am in agreement with. My modified definition is meant to highlight the downstream activity that should be significant for First Amendment scholars and lawyers).Communicative infrastructure, as well as the absence of state censorship, is necessary for free speech, including public opinion formation and debate.18 18. Jack M. Balkin, Digital Speech and Democratic Culture: A Theory of Freedom of Expression for the Information Society, 79 N.Y.U. L. Rev. 1, 52–55 (2004). Many democratic constitutions restrict the regulation of the press because the press is a type of communications infrastructure, and such infrastructure supports public discourse.19 19. Post, supra note 3.

Widely used communicative products, such as Facebook and TikTok, have additional infrastructural characteristics. They are embedded in other social arrangements and technologies, they invisibly support tasks, they reach beyond single events and practices, new participants acquire familiarity with them by using them, they have a standardizing influence on other infrastructures and tools, and they become visible upon breakdown.20 20. Susan Leigh Star & Karen Ruhleder, Steps toward an Ecology of Infrastructure: Design and Access for Large Information Spaces, 7 Information Systems Research 111 (1996) republished in Bowker et. al. (ed.) Boundary Objects and Beyond, 380-381(2015); Cohen (2019), supra note 11, at 235. These elements are also increasingly found in chatbots that rely on large language models and AI-based creation tools embedded in social media apps.

III. Infrastructure is Different from Ideas

As Jack Balkin has pointed out, the First Amendment prevents state interference with speech, but infrastructure is necessary to ensure that individuals have the freedom and ability to participate in public discourse.21 21. Jack Balkin, Cultural Democracy and the First Amendment, 10 Northwestern Law Review 1053, at 1060, 1078 (2016). Consider the infrastructure required for public discourse: to access materials and inspiration, to consult others, and to circulate ideas and conclusions. Communicative infrastructure—extending from the press to generative AI—is critical to maintaining public discourse through its structural role in disseminating information.22 22. See Rahman, supra note 11 at 1670–1. This is different from and at least as important as the information itself. Along with the right to access ideas, one needs the means to access ideas.23 23. See generally James Grimmelmann, Listeners’ Choices, 90 U. Colo. L. Rev. 365 (2019).

Furthermore, communicative infrastructure exerts an influence beyond its own systems.24 24. Frischmann, supra note 3 at 337-9. When the state restricts the speech of, say, a significant university or the operations of a major social media platform, more people are affected than those directly involved with the university or platform. Universities and platforms, in different ways and for different reasons, make certain associations and speech salient. Major social media platforms influence what stories the traditional media carry, how politicians campaign, and even how democratic institutions share information with the public.25 25. Joan Donovan & danah boyd, Stop the Presses? Moving From Strategic Silence to Strategic Amplification in a Networked Media Ecosystem, 65 American Behavioral Scientist 333 (2021); Robyn Caplan & danah boyd, Isomorphism through Algorithms: Institutional Dependencies in the Case of Facebook, 5 Big Data & Society 1 (2018); Madison Malone Kircher, How Harris, Walz and Trump Are Finding Their Way to Your TikTok Feed, The New York Times, Aug. 30, 2024, https://www.nytimes.com/2024/08/30/style/tiktok-creators-harris-walz-trump.html; Liam Scott, White House to Open Media Access to Podcasters, Influencers, Voice of America (Jan. 28, 2025), https://www.voanews.com/a/white-house-to-open-media-access-to-podcasters-influencers/7953761.html. Other platforms change their practices to compete.26 26. Sarah Jackson, Mark Zuckerberg Told Meta Employees to Zero in on Video Because They’re up against an “unprecedented Level of Competition” from TikTok, Report Says, Business Insider, https://www.businessinsider.com/mark-zuckerberg-meta-faces-unprecedented-competition-from-tiktok-focus-reels-2022-2 (last visited Apr. 7, 2025).

The infrastructural lens illuminates important elements of the relationship between speech, association, and communicative online apps, including social media platforms and AI-powered chatbots. Communicative online apps play a dual role, enabling communicative relationships at an unprecedented scale while also influencing and shaping them in ways that may impede democracy. For example, social media platforms can change what content is available, ranging from single items of content to individual accounts and even whole feeds, and can use removal, amplification, downranking, and shadow banning to shape the kinds of content that are available.27 27. Cohen, supra note 11 at 40 (Collective or[external] harms that inhere in patterns of activity manifesting at scale can be particularly difficult to name, understand, and counteract, and this is doubly true for collective harms entrenched infrastructurally); See also Rahman, supra note 11 at 1672. Through repeated nudges, platforms also influence who is put into conversation with whom.

A single entity can play a greater infrastructural role if it dominates the market, as Google does. When Google’s products are unavailable, the public notices.28 28. Stacy Liberatore, Google Hit with Outage That Plagued Thousands of Users, Mail Online (2024), https://www.dailymail.co.uk/sciencetech/article-13735777/Google-Search-engine-YouTube-outage.html. That is a feature of infrastructure—it is seamless and invisible when it works and visible when it stops working. Many people rely on Google. Courts are increasingly finding that this reliance gives Google too much infrastructural power.29 29. See, e.g., United States v. Google LLC 778 F.Supp.3d 797 . See also Rahman, supra note 11; Lina Khan, The Separation of Platforms and Commerce, 119 Colum. L. Rev. 973 (2019); Amy L. Stein, Rejecting Public Utility Data Monopolies, 112 Cal. L. Rev. 1321 (2024). This is important because not all forms of communicative infrastructure may serve democratic values. Non-state actors can also erode infrastructure, which is why structural regulation, such as anti-trust laws, is necessary to prevent private capture of communicative infrastructure that would stunt the diversity of opinion and group formation.

Social media has a powerful effect on public discourse because it steers communicative relationships. Recognizing this associational power and interpreting the First Amendment to account for it is critical for a healthy public sphere. Up to and including the era of broadcast media, regulating media to maximize its benefits to democracy was seen as consistent with First Amendment values, but the standards changed before the social media cases reached the Supreme Court.30 30. See Twitter Inc. v. Taamneh, 598 U.S. 471 (2023); Murthy v. Missouri, 603 U.S. 43; Moody v. NetChoice, LLC, 144 S.Ct. 2383 (2024); TikTok Inc. v. Garland, 604 U. S.; Jack M. Balkin, Moody v. NetChoice: The Supreme Court Meets the Free Speech Triangle, Sup. Ct. Rev. 1 (2024); Douek and Lakier, supra note 11; Genevieve Lakier, The TikTok Ban and the Limits of the First Amendment, LPE Project (Jan. 22, 2025), https://lpeproject.org/blog/the-tiktok-ban-and-the-limits-of-the-first-amendment/. Veering away from the idea that communicative infrastructure can be regulated to enable and protect First Amendment interests has left policymakers and judges without theories of how to treat foreign communicative infrastructure. Specifically, treating ideas as medium-agnostic and treating media as ownership-agnostic ignores how far ownership affects the manner in which infrastructures shape speech and association. If foreign ideas are valuable, so are foreign ways of relating and communicating.

IV. The Special Role of Foreign Communicative Infrastructure

Apps shape and are shaped by the communities they engage with, the markets they compete in, the organizations that build them, their access to resources, and local laws, among other things.31 31. Id.; Zeyi Yang, How Chinese AI Startup DeepSeek Made a Model That Rivals OpenAI, Wired, https://www.wired.com/story/deepseek-china-model-ai/ (last visited Feb. 1, 2025); Susan Leigh Star & Karen Ruhleder, Steps toward an Ecology of Infrastructure: Design and Access for Large Information Spaces, 7 Information Systems Research 111, 380-381 (1996). Tellingly, WhatsApp was co-founded by a man whose Soviet origins taught him the importance of encrypted communication.32 32. David Rowan, The inside Story of Jan Koum and How Facebook Bought WhatsApp, Wired (May 1, 2018), https://www.wired.com/story/whats-app-owner-founder-jan-koum-facebook/. The Chinese company DeepSeek’s large language models were shaped by resource constraints; the company’s engineers innovated in response to limited access to the powerful new generation of Nvidia chips that power generative AI.33 33. Yang, supra note 31. Local dynamics in China trigger innovation in different directions. This produced ByteDance’s platform Douyin’s evolving business practices and addictive algorithm for sharing and watching short videos. China has a highly competitive short-video market. ByteDance’s platform for international users, TikTok, has successfully tailored learnings from Douyin for international markets.34 34. D. Bondy Valdovinos Kaye, Xu Chen & Jing Zeng, The Co-Evolution of Two Chinese Mobile Short Video Apps: Parallel Platformization of Douyin and TikTok, 9 Mobile Media & Communication 229 (2021) 230-1.

As the spread of the short-video format and TikTok’s practices to engage users to other popular communicative platforms shows, foreign innovation can revitalize the infrastructural practices of American communicative apps.35 35. Mark Zuckerberg Is Blowing Up Instagram to Try and Catch TikTok, Bloomberg.com, May 25, 2022, https://www.bloomberg.com/news/features/2022-05-25/facebook-copies-tiktok-app-to-make-instagram-cool-to-teens; Amani Bayo, How Social Media Influencers Make Money on TikTok, Instagram and YouTube, The Columbus Dispatch, https://www.dispatch.com/story/entertainment/local-celebrity/2025/06/22/how-to-make-money-on-social-media-tiktok-instagram-youtube/82641137007/ (last visited July 24, 2025). Even American social media users who have never engaged with any TikTok videos are affected by changes that American platforms make as they innovate to keep up with ByteDance.36 36. Liv McMahon, Instagram Revamps Algorithm in Battle for TikTok Creators, BBC, May 3, 2024, https://www.bbc.com/news/articles/cmm3yn4pr17o; Emma Job, Instagram Is Considering Creating a Reels App to Compete with TikTok, Tech Edition (Feb. 27, 2025), https://www.techedt.com/instagram-is-considering-creating-a-reels-app-to-compete-with-tiktok. The Chinese company is a major competitive presence in a highly concentrated market and triggers changes in the social media ecosystem that can benefit public discourse as well as individuals.37 37. Social Media Fact Sheet, Pew Research Center (Nov. 13, 2024), https://www.pewresearch.org/internet/fact-sheet/social-media/ (showing a sharp rise in the percentage of adults who use TikTok). Admittedly, these changes such as increased addictiveness can also be harmful, but it is difficult to separate out the harm from the benefits just as it is difficult to separate bad ideas from good ones.

Forcing changes in ownership and making foreign social media platforms American-owned changes the nature of communicative apps in ways that affect speech.38 38. It is too early to be certain but the TikTok deal may have avoided this outcome by licensing the Chinese algorithm which may offer it access to Chinese updates but the licensing has been criticized by those who are opposed to Chinese ownership of social media platforms widely used in the United States. See Associated Press, Larry Ellison’s Oracle Set to Spearhead U.S. Oversight of TikTok Algorithm, Fortune, https://fortune.com/2025/09/23/oracle-tiktok-deal-us-security-algorithm-silverlake-murdoch-dell-trump-2025/ (last visited Jan. 1, 2026); Bobby Allyn, TikTok Signs Deal to Give U.S. Operations to Oracle-Led Investor Group, NPR, Dec. 18, 2025, https://www.npr.org/2025/12/18/nx-s1-5648844/tiktok-deal-oracle-trump. Social media platforms’ design and content choices have an enormous impact on communication. For example, platforms enable and classify participation differently. TikTok’s algorithm recommends third-party content based on how long users engage with videos and react to them.39 39. Yulun Ma & Yue Hu, Business Model Innovation and Experimentation in Transforming Economies: ByteDance and TikTok, 17(2) Management and Organization Review, 382-388 (2021) at 385 (“TikTok touted the algorithm as computer vision to extract and categorize visual information that relies on users’ watching history and engagement patterns to serve users’ interests. In short, TikTok controls your menu of entertainment by observing your reactions to each past video. Therefore, TikTok users don’t need to think and search for the videos but are fed personal preference-based videos, which is a crucial part of TikTok”); Compare with Angela Yang, Instagram and Twitch roll out new TikTok-like short-form video discovery features, NBC News, May 1, 2024 instagram-tiktok-roll-new-tiktok-short-form-video-discovery-features-rcna150231.html (“Instagram announced Tuesday a tweak in its discovery algorithm that would amplify smaller creators through recommendations, which show users posts and reels from accounts they do not already follow through an algorithm that tracks and predicts their interests.”). Those who access this infrastructure have a different communicative experience from those who use other platforms.

A new owner of a communicative app can order communication and association in new ways. The most widely used social media and, increasingly, AI platforms follow an infrastructural model that Julie Cohen describes as “platformized” infrastructure.40 40. See Cohen (2023), supra note 11 at 4, 26; Cohen (2019), supra note 11 at 41. This model enables greater centralized control than is typical for public infrastructure such as a public library system.41 41. Cohen (2019), supra note 11 at 41–2; See also Laura J. Neumann & Susan Leigh Star, Making Infrastructure: The Dream of a Common Language, Proceedings of the 4th Biennial Participatory Design Conference (1996), https://ojs.ruc.dk/index.php/pdc/article/view/153. Platformization places greater control in platform owners’ hands, permitting their ideological orientations and their susceptibility to markets, law, and political power to shape platforms’ decisions about what content and communication they permit and privilege.

Owners’ influence can be relatively subtle on social media platforms and AI-driven communicative platforms because they engage in what I am going to call ‘algorithmic content mediation,’ to distinguish it from editing and curation, both of which imply more precision of control over the output. Newspapers know exactly what they will publish, and their owners have to tell (human) editors what they cannot publish.42 42. Richard Luscombe, Washington Post Opinion Editor Departs as Bezos Pushes to Promote ‘Personal Liberties and Free Markets,’ The Guardian, Feb. 26, 2025, https://www.theguardian.com/media/2025/feb/26/jeff-bezos-washington-post-opinion. In contrast, online platforms can use filters and prioritize some kinds of content over others, but they rarely know exactly what will emerge given the speed and scale of speech online.43 43. Jack M. Balkin, Free Speech versus the First Amendment, 70 UCLA L. Rev. 1206 (2023); Paul Gowder, The Networked Leviathan 46–48, 126–137 (2023).  For example, when Elon Musk started changing Twitter’s (now X’s) algorithms, a menswear writer called Derek Guy ended up being promoted all over the platform for reasons that remain unclear. This is why it is helpful to think of algorithmic content mediation as “deal[ing] in probabilities, error rates, costs, and benefits.”44 44. Balkin, Free Speech versus the First Amendment,70 UCLA Law Review 1206 at 1249.

V. Foreign Ownership and Communicative Infrastructure

By stressing the value of foreign communicative platforms, I am not implying that they should not be regulated.45 45. See Sitaraman, supra note 5 at 1077–8. This essay confines itself to demonstrating that foreign apps contribute more value to the American public sphere than regulatory conversations acknowledge. Regulating foreign ownership of communicative infrastructure should involve assessing how far foreign political and cultural influences enrich American public discourse. Foreign apps, with their distinct influences, offer new ways for people to engage with each other. Infrastructural apps like major social media platforms also evolve continuously. This ought to be taken into account by the Supreme Court when it assesses how far regulating foreign apps burdens speech.

TikTok Inc. v. Garland offers an illustration of what is lost when judges do not account fully for the burden on speech.46 46. TikTok Inc. v. Garland, 604 U. S. __. The Supreme Court acknowledged that the First Amendment rights of American users were implicated in the use of TikTok. However, it underestimated the extent and nature of the relationship between these rights and major foreign platforms like TikTok by failing to account for TikTok’s infrastructural role, and how this role would be affected by transferring TikTok to American owners. In the Supreme Court’s defense, TikTok’s arguments in the case did not highlight its infrastructural role or influence on public discourse and confined itself to discussing how its users were affected.

TikTok enables Americans’ First Amendment rights in three significant, under-considered ways: (1) content from TikTok circulates to other platforms; (2) TikTok, as a foreign platform, offers Americans unique ways to communicate and engage; and (3) TikTok’s practices trigger changes in competing platforms.47 47. See for example, McMahon, supra note 36; Job, supra note 36. Thus, access to Chinese (and other foreign) apps supports Americans’ First Amendment rights.

Lawmakers, judges, lawyers, and others described TikTok as affecting only the rights of its 170 million registered American users rather than accounting for the entire burden on speech.48 48. See TikTok Inc. v. Garland, 604 U.S. __. The enormity of this burden becomes clear when one considers that TikTok affects the speech of more Americans than just its users, who make up a third of all American adults and over half of Americans under 30.49 49. Kirsten Eddy, 8 Facts about Americans and TikTok, Pew Research Center (Dec. 20, 2024), https://www.pewresearch.org/short-reads/2024/12/20/8-facts-about-americans-and-tiktok/. TikTok permits these people to engage with more than a billion people across the world.50 50. Andreas Schellewald, Understanding the Popularity and Affordances of TikTok through User Experiences, 45 , Media, Culture and Society 1568 (2023). It triggers changes in the cultural industries and the social media sector, as other companies, media, and content creators react to TikTok’s innovations.51 51. See Jackson, supra note 29.

As I was writing this essay, President Trump approved a TikTok transfer of ownership agreement.52 52. David Shepardson, US Treasury’s Bessent Says China Has Approved TikTok Transfer Deal, Reuters, Oct. 30, 2025, https://www.reuters.com/world/china/china-us-trade-deal-could-be-signed-next-week-us-treasurys-bessent-says-2025-10-30/. The arrangement did not take the form that most commentators and lawyers—and arguably the Supreme Court—had envisioned it might: namely, that the algorithm would be transferred when the ownership changed.53 53. See Alan Rozenshtein, Lessons Learned from the TikTok Saga, Lawfare (2025), https://www.lawfaremedia.org/article/lessons-learned-from-the-tiktok-saga ("As someone who analyzed and defended the law on both legal and policy grounds from the very beginning, I should be happy with this outcome. And yet I am anything but. I have serious concerns about whether this deal genuinely addresses the national security risks—the opacity of any algorithmic "licensing" arrangement makes verification impossible"). Instead, the TikTok agreement arranged for ByteDance to license the content recommendation algorithm to TikTok’s new owners.54 54. See TikTok’s Algorithm to Be Licensed to US Joint Venture Led by Oracle and Silver Lake, AP News (Sept. 22, 2025), https://apnews.com/article/trump-tiktok-china-d5d8a1d56b5185778536874d7fc1ee62; Rozenshtein, supra note 53. However, I am going to discuss the Supreme Court’s decision in line with the initial expectations that ownership of the algorithm would change, as that allows an examination of the broader question of how ownership affects digital communicative infrastructure.

New owners have a visible impact on the development of platformized apps for the reasons I highlighted in the previous section. This is evident from Meta’s goals for WhatsApp.55 55. Parmy Olson, Exclusive: WhatsApp Cofounder Brian Acton Gives The Inside Story On #DeleteFacebook And Why He Left $850 Million Behind, Forbes.com 4 (2018); Alex Hern & Alex Hern UK technology editor, Twitter Usage in US ‘Fallen by a Fifth’ since Elon Musk’s Takeover, The Guardian, Mar. 26, 2024, https://www.theguardian.com/technology/2024/mar/26/twitter-usage-in-us-fallen-by-a-fifth-since-elon-musks-takeover.Although WhatsApp’s co-founders chose to sell the app to Meta (then Facebook), news reports suggest that they were unhappy with the new owners’ design choices.56 56. Olivia Solon, WhatsApp CEO Jan Koum Quits over Privacy Disagreements with Facebook, The Guardian, Apr. 30, 2018, https://www.theguardian.com/technology/2018/apr/30/jan-koum-whatsapp-co-founder-quits-facebook; Olson, supra note 55. Similarly, Elon Musk’s influence on X, previously known as Twitter, offers an extremely visible example of the impact of ownership changes on communicative platforms, including social media.57 57. See Kate Conger & Ryan Mac, Character Limit (2024), https://www.penguinrandomhouse.com/books/737290/character-limit-by-kate-conger-and-ryan-mac/; Antonio Pequeño IV, Meta’s Threads, Bluesky Gain Users As Musk’s X Dips, Forbes, https://www.forbes.com/sites/antoniopequenoiv/2025/03/25/elon-musks-x-sees-popularity-slide-after-election---while-threads-bluesky-makes-gains/ (last visited June 24, 2025); Dan Milmo & Dan Milmo Global technology editor, ‘Musk Destroyed All That’: Twitter’s Business Is Flailing after a Year of Elon, The Guardian, Oct. 27, 2023, https://www.theguardian.com/technology/2023/oct/27/elon-musk-x-twitter-takeover-revenue-users-advertising; Aisha Counts, Elon Musk Is a ‘Free Speech Absolutist,’ Except at Work, Bloomberg.com, Sept. 14, 2023, https://www.bloomberg.com/news/newsletters/2023-09-14/elon-musk-says-he-s-pro-free-speech-but-fired-twitter-staff-for-comments. A change in ownership means new priorities, values, and goals inform an app’s evolution, which implies that changes in ownership transform communication apps.

Access to a diversity of communicative apps, including foreign apps, is critical to Americans’ right to explore new ways of engaging with each other. The major Chinese platforms are currently the only platforms that offer Americans an alternative to discourse ordered by what is increasingly called the American “tech oligarchy.”58 58. See Julie E. Cohen, Oligarchy, State, and Cryptopia (abstract), dli-cornell-tech, https://www.dli.tech.cornell.edu/seminars/oligarchy%2C-state%2C-and-cryptopia (draft on file); See also Chris Hayes: The Dystopian Vision Trump and His Billionaire Allies Have for the U.S., MSNBC.com (Jan. 8, 2025), https://www.msnbc.com/top-stories/latest/trump-musk-zuckerberg-bezos-oligarchy-rcna186818. Although there are good reasons to regulate foreign apps (and American apps that threaten national security), there is no reason for the regulation to be so expansive that it blocks Americans’ access to them instead of regulating their harmful effects or tendencies narrowly. In this regard, the European Union’s approach to foreign platforms might have been the better approach for American policymakers to adopt. Robert Post has argued that the European Union should treat Google, which is foreign-owned in the EU, as “an essential component of the communicative infrastructure necessary to sustain the public sphere” because of its status as a significant online platform.59 59. Robert C. Post, Data Privacy and Dignitary Privacy: Google Spain, the Right to Be Forgotten, and the Construction of the Public Sphere, 67 Duke L.J. 981, 1016, 1070-71 (2018). This argument could also apply to foreign communicative platforms in the United States: They enable Americans’ First Amendment rights. However, the dominant social media platforms in the United States and most of the English-speaking world have historically been American. This has left American regulatory discourse unprepared for the constitutional questions that foreign platforms raise. Policy aside, the Supreme Court would have been better able to appreciate the scale of speech rights at stake if the problem in TikTok Inc. v. Garland had been presented as a right to access communicative infrastructure. Framed correctly, the TikTok case might have been the Lamont of modern media.

VI. Adapting Lamont for Web-based Media

Lamont v. Postmaster General, a 1965 Supreme Court case, addressed whether an American citizen’s First Amendment rights were violated when the U.S. Post Office detained his copy of the Chinese news magazine Peking Review in 1963. In Lamont, the Supreme Court ruled that controlling the free flow of mail to the American public violates the First Amendment, just as controlling the free flow of ideas does.60 60. 381 U.S. 301, 306 (1965) [“just as the licensing or taxing authorities in the Lovell, Thomas, and Murdock cases sought to control the flow of ideas to the public, so here federal agencies regulate the flow of mail”]; See also Martin v. City of Struthers, 319 U.S. 141 (1943). The law that was struck down required recipients of “communist propaganda” to take certain steps to receive it. This was an impediment (as opposed to an obstacle) that the Supreme Court interpreted as a violation of the First Amendment. Although Lamont is considered a case about the rights of American listeners, it could be framed as addressing the regulation of American communicative infrastructure, specifically, the U.S. Postal Service.61 61. See, e.g., Wang, supra note 1; Grimmelmann, supra note 23 at 399–400.

Recognizing listeners’ rights was pathbreaking at the time and remains a relatively underdeveloped thread of First Amendment jurisprudence and scholarship.62 62. Wang, supra note 1. The ideal progression from Lamont would have been if First Amendment jurisprudence discussed Americans’ right to access foreign radio and television. In the period before Lamont and the world wars, foreign-owned radio was restricted in the United States.63 63. Sitaraman, supra note 5 at 1115–6. The concern at the time, although it included foreign propaganda, was primarily that the spectrum was limited and that foreign entities might find a way to run signal interference with American military communication.64 64. Adeno Addis, Who’s Afraid of Foreigners? The Restrictions on Alien Ownership of Electronic Media, 32 Colum. Hum. Rts. L. Rev 133, 147. The scarcity doctrine generally allowed for regulating broadcasting infrastructure in ways that were directly linked to the technological constraints of the limited spectrum.65 65. Turner Broadcasting System, Inc. et al. v. FCC I, 512 U.S. 622 (Supreme Court of the United States 1994). Legal norms regulating foreign ownership of broadcasting originated in these technological constraints and not in what might have been a helpful discussion of how threats to national security should be measured against Americans’ right to access foreign infrastructure.

TikTok Inc. v. Garland illustrates why thinking in terms of communicative infrastructure matters. TikTok Inc. v. Garland was about Americans’ access to foreign communicative infrastructure, even though it was never framed in this way for the court. Access to infrastructure affects access to ideas, and offers a sharper framing of what the justices might have been alluding to when they acted to protect the free flow of mail in Lamont.66 66. Lamont v. Postmaster General, 381 U.S. 301, at 306. Freedom of speech depends on the technological infrastructure that enables widespread democratic participation and supports the system of free expression.67 67. Jack M. Balkin, Digital Speech (2004) supra note 18 at 6. Social media audiences participate even through what looks like passive consumption. For example, watching a video about climate change is engagement because when people watch these videos, they tell the algorithm about their own likelihood, as well as their demographic group’s likelihood, of watching other videos like it. On TikTok, Lamont’s passive listener becomes a participant.

Modern communicative infrastructure is complicated because, unlike the postal service, it is privately owned. Nevertheless, protecting communicative infrastructure, including online platforms, is necessary to protect First Amendment values. The First Amendment cannot protect free speech values without the institutions, infrastructure, and industries that promote these values.68 68. Jack M. Balkin, Free Speech (2023) supra note 43. First Amendment law protects some kinds of communicative infrastructure, like the press and broadcast media, but fails to account for forms of communicative infrastructure that technology and online platforms create, although social media platforms are the dominant online communication model.

VII. Restricting TikTok 

The events leading to the TikTok restriction may be familiar to many readers. President Trump initiated the TikTok ban during his first term, and President Biden’s administration sealed it in place with the Protecting Americans from Foreign Adversary Controlled Applications Act (PAFACA).69 69. See Anupam Chander, Trump v. TikTok, 55 Vanderbilt Journal of Transnational Law 1145, 1148–56 (2022); Anupam Chander & Paul Schwartz, The President’s Authority over Cross-Border Data Flows, 172 U. Pa. L. Rev. 1989, 1990–1 (2024). PAFACA targets TikTok and potentially any other “foreign adversary controlled applications” by permitting the executive to designate them for restriction.70 70. 15 U.S.C. § 9901. Both administrations treated TikTok as presenting national security concerns because it was Chinese owned and controlled and therefore subject to the Chinese government, which they feared could influence data collection and algorithmic content targeting.

The D.C. District Court, before which the statute was challenged, deferred to the legislature’s and executive’s characterization of the problem, which can be summarized as follows: (1) China is a serious cyber espionage threat to the United States; (2) China has made significant efforts to gather data about Americans for intelligence purposes; (3) China has gathered data from Chinese companies as well as foreign companies in which it makes investments or from which it purchases data sets; (4) Chinese laws require Chinese companies to provide China with full access to their data for criminal and security investigations; and (5) Chinese companies are not independent in the way in which American companies are.71 71. TikTok v. Garland, 122 F.4th 930, 953–4 (2024).Additionally, the assistant director of the FBI asserted that (6) China tries to pre-position companies in the US that it can co-opt later and (7) engages in influence campaigns to counter and suppress opinions it objects to about China.72 72. Id. Additionally, the US government concluded, based on TikTok Global’s responsiveness to Chinese demands to censor content, that (8) TikTok would comply if the Chinese government asked it to manipulate content in the United States, especially since China maintains “a powerful Chinese Communist Party committee embedded in ByteDance,” (internal quotation marks omitted) and TikTok is reliant on and integrated with ByteDance.73 73. Id. at 954–5. Up to the point of the district court hearings, content-related concerns featured prominently among the justifications for banning TikTok.

In contrast, the Supreme Court focused on TikTok’s data collection, eliding the concerns about content and ignoring PAFACA’s regulation of expressive activity. Focusing exclusively on its data collection goal, the court argued that Congress would have passed the law even if it were not concerned about manipulative content.74 74. TikTok Inc. v. Garland, 604 U.S. __; Genevieve Lakier, The TikTok Ban and the Limits of the First Amendment, LPE Project (Jan. 22, 2025), https://lpeproject.org/blog/the-tiktok-ban-and-the-limits-of-the-first-amendment/. After framing data collection as the law’s goal, the Supreme Court argued that PAFACA regulated conduct rather than speech and applied the intermediate scrutiny test, not the strict scrutiny test.75 75. TikTok Inc. v. Garland, 604 U.S. __. The Court concluded that PAFACA, as applied to TikTok, satisfied the intermediate First Amendment scrutiny test because its objective—which the justices saw as preventing a foreign adversary, China, from leveraging its control of ByteDance to capture TikTok users’ personal data—qualified as an important government interest.76 76. Id. The justices also decided that PAFACA did not burden substantially more speech than necessary.77 77. Id.

VIII. Why TikTok Inc. v. Garland is Troubling

The Supreme Court’s decision should concern everyone interested in speakers’ rights, listeners’ rights, healthy public discourse, and the rights of freedom of thought and association. The justices’ conception of social media platforms led them to underestimate the potential impact of divestment on First Amendment interests. The Supreme Court does not appear to have carefully considered how to protect expressive or associational activity in TikTok Inc. v. Garland: It treated TikTok as static, ignoring its role as an evolving, ubiquitous infrastructure. In addition to concluding that the statute did not burden substantially more speech than necessary, the Supreme Court argued that the statute’s “TikTok-specific provisions” requiring divestment of the company are facially content neutral, in the context of whether strict scrutiny applies.78 78. Id. If the question before the justices had been framed as one of access to foreign infrastructure, different considerations might have been weighed against the national security interests animating PAFACA.

Major social media platforms, including TikTok, exhibit infrastructural features: they can be consumed non-rivalrously, enable downstream productive activity, and act as input into a wide range of goods and services.79 79. See Frischmann, supra note 3 at 61. Their ownership plays a role in how they do these things. It is possible that the leasing of the TikTok algorithm from Douyin will ensure that TikTok’s unique infrastructural role is not lost. However, the Supreme Court did not anticipate or require this outcome. In other words, it did not consider how Americans’ First Amendment interests are served by access to foreign communicative infrastructure.

The Supreme Court’s assessment of whether speech was affected and to what degree was based on an outdated understanding of how web-based platforms and Chinese platforms affect Americans’ First Amendment interests. Communicative platforms often offer entirely new ways of communicating. Social media platforms shape how communication is “transmitted, received, and recirculated,” and the “experiences and perceptions of what is ordinary and natural in the linked realms of private and public interconnection and sensemaking.”80 80. Cohen (2023), supra note 11 at 16. They also shape which people engage with each other and how people interact with bots. By affecting online association, they influence communication.81 81. For a discussion of the relationship between freedom of speech and freedom of association, see generally Ashutosh Bhagawat, Associational Speech, 120 Yale L.J. 978 (2011).

The infrastructural properties of each platform are distinct. Some platforms privilege content that elicits reactions from others, and other platforms (e.g., Bluesky) avoid rewarding reactions, choosing to treat the sharing of someone else’s post in exactly the same way as annotated sharing of the same post.82 82. See Wes Davis, Here’s some cool stuff you can do with Bluesky, The Verge, Nov. 16, 2024, https://www.theverge.com/24295933/bluesky-social-network-custom-how-to (about BlueSky’s simple chronological feed). Platforms also construct associations differently (although their choices may affect those of their competitors); TikTok’s method of profiling users has influenced Meta.83 83. McMahon, supra note 36. Safeguarding infrastructural diversity does not imply that platforms cannot be restricted or regulated, but shows that their differences play an important First Amendment role that should be considered while regulating them.

IX. Conclusion

The value of Americans’ right to access communicative infrastructure, including foreign communicative infrastructure, is under-considered in legal debate about communicative apps. Newcomers to the social media and artificial intelligence industries struggle to break through and are almost always purchased by the dominant firms, as WhatsApp and Instagram were.84 84. The AI industry also has a higher entry barrier because of the data and computational power required to train models. See Jennifer Cobbe, Michael Veale & Jatinder Singh, Understanding Accountability in Algorithmic Supply Chains1186 (2023), https://doi.org/10.1145/3593013.3594073. This leaves online communicative infrastructure very vulnerable to capture by private power.85 85. Tim Wu, The Master Switch 310 (2010). Focusing on citizens’ access to multiple sources of information and modes of engagement means that foreign communicative apps will remain an important source of alternative information, engagement, and competition. The judiciary’s willingness to defer to the legislative and executive branches’ characterization of communicative infrastructures as national security threats without careful consideration of their democratic value substantially undermines the First Amendment because of the scale of the speech and association affected.

If foreign-owned apps are easily restricted, communicative infrastructure will be left largely in the hands of a few American companies. Although these companies are not controlled by foreign states, they do have an interest in foreign markets and competing for profits. There is no reason to assume that they will never present or enable a national security threat. Using foreign ownership or control as a proxy for determining such a threat avoids the question of how communicative infrastructure should be regulated to mitigate national security threats and promote First Amendment values. Scholars have already proposed strategies, including fiduciary responsibility, interoperability, and anti-trust regulation, that might mitigate concerns about social media infrastructure.86 86. Cohen, supra note 11 at 16, 19. Layering additional regulations to address national security concerns is a more effective way to address concerns about foreign states than barring foreign-owned or foreign-controlled social media. Until American structural regulation improves, it is likely that apps capable of competing with dominant US communicative apps will only grow significantly in foreign countries with protected markets

Because non-US apps are shaped by foreign political, cultural, and economic forces, they create an environment for online engagement that American companies cannot create. They offer Americans what American companies have offered other publics in other countries for years: new ways to engage with ideas, with each other, and with people across borders. As communicative infrastructure, these apps are significant. To compel them to divest is to remove the quality that makes them unique: that their design reflects a distinct foreign influence.

Acknowledgments

I thank Maria Angel Arango, Jack Balkin, Katherine G. Bass, Liang Chen, Julie Cohen, Yuval Erez, Akriti Gaur, Jennifer Heinrichson, Margaret Hu, Jameel Jaffer, Shira Minsk, Michael McGovern, John Langford, Anat Leshnick, Lyrissa Lydsky, Yuping Lin, Dr. Margaret A. Oppenheimer and Cambridge Proofreading, John Peters, Robert Post, Alan Rozenshtein, Nishant Shah, David Schulz, Kalindi Vora, George Wang, Caleb Yong and Ethan Zuckerman. All mistakes are mine.

© 2026, Chinmayi Arun.

Cite as: Chinmayi Arun, The Right to Access Foreign Communicative Infrastructure, 26-4 Knight First Amend. Inst. (May 1, 2026), https://knightcolumbia.org/content/the-right-to-access-foreign-communicative-infrastructure [https://perma.cc/TQ6Q-48TB]. 

The following can be attributed to Jameel Jaffer, executive director at the Knight First Amendment Institute at Columbia University:

“The FCC has no authority to cancel broadcasters’ licenses because of their perceived political views. But this isn’t just about the rights of Disney and ABC. President Trump is trying to consolidate control over what Americans see and hear on the radio, television, and social media. If he gets his way, we’ll have only government-aligned media organizations that broadcast only government-approved news and commentary. It would be difficult to imagine an outcome more corrosive to democracy or more offensive to the First Amendment.”

For more information, contact: Adriana Lamirande, adriana.lamirande@knightcolumbia.org 

The following can be attributed to Xiangnong (George) Wang, staff attorney at the Knight First Amendment Institute at Columbia University:

“We welcome the Mayor’s veto of legislation that would have charged the NYPD with setting the rules around lawful political protest near thousands of schools and educational facilities across the city. This measure risks chilling and even criminalizing a wide range of speech the First Amendment protects, particularly at some of the city’s most vital sites for public discourse. Governments shouldn’t entrust law enforcement with determining the limits of political expression.”

For more information, contact: Adriana Lamirande, adriana.lamirande@knightcolumbia.org

The following can be attributed to Carrie DeCell, senior staff attorney and legislative advisor at the Knight First Amendment Institute at Columbia University: 

“The State Department is excluding tech researchers from the United States because of their constitutionally protected work, and this newly disclosed memo only underscores the unbounded scope and unconstitutionality of the policy. That the Trump administration claims its censorial policy is intended to combat censorship is the height of doublespeak.”

In March, the Knight Institute and Protect Democracy filed a lawsuit on behalf of the Coalition for Independent Technology Research (CITR) challenging the constitutionality of the policy. The complaint argues that the policy violates the First Amendment and chills independent research and reporting about social media and other internet platforms.

Read more about the case here.

For more information, contact: Adriana Lamirande, adriana.lamirande@knightcolumbia.org 

Spyware tools like Pegasus, developed by NSO Group, enable covert access to journalists’ phones, communications, and personal data, often through “zero-click” attacks that require no user interaction. Farid Johnson emphasized that these technologies endanger journalists and their sources while disrupting newsroom operations, chilling investigative reporting, and limiting the public’s access to information.

Farid Johnson also highlighted the legal barriers spyware victims face in U.S. courts, noting that procedural hurdles have kept many cases from reaching the merits. She urged Congress to amend the Computer Fraud and Abuse Act to clarify that U.S. courts should hear claims involving spyware attacks that exploit U.S.-based technology infrastructure, ensuring a meaningful pathway to redress.

The hearing also featured testimony from Sergio Arauz, deputy editor-in-chief of El Faro and a plaintiff in Dada v. NSO Group—the Institute’s lawsuit on behalf of journalists targeted with Pegasus spyware. He described escalating repression under El Salvador’s ongoing state of exception, including surveillance, exile, and the criminalization of journalism. His testimony underscored the real-world consequences of the dynamics Farid Johnson identified, linking the use of spyware to broader patterns of democratic erosion and the silencing of dissent.

Read Farid Johnson’s full statement submitted for the record, here. 

Read Sergio Arauz’s full testimony, here. 

Watch a recording of the full hearing below.

Around the world, leaders like Recep Tayyip Erdoğan in Turkey and Viktor Orbán, the former prime minister of Hungary, have used legal systems, rules of law, and institutional practices to consolidate power, restrict dissent, and hollow out democratic accountability from within. That pattern is becoming more visible in the United States, where mounting political pressure on courts, lawyers, and legal institutions is raising urgent questions about the role of the legal profession in moments of democratic crisis. 

“Lawyering Without Law,” a bi-weekly podcast from the Knight Institute, interrogates the unique and important role that lawyers play in defending democracy, or in facilitating the slide into authoritarianism. Hosted by Knight Institute Senior Fellow and Columbia Law Professor Madhav Khosla and the Knight Institute’s Research Director Katy Glenn Bass, the series brings together scholars, litigators, and practitioners to explore these dynamics across historical and contemporary contexts. Drawing on global examples of democratic backsliding, each episode connects these developments to the United States and outlines what is at stake for the legal profession and for democracy itself. 

Read more about Khosla’s research project with the Knight Institute examining the crucial role that lawyers can play in preserving democratic freedoms and institutions here.

“Lawyering Without Law” is available on Apple, Spotify, and wherever you get podcasts.

Listen, subscribe, and leave a review. We’d love to know what you think.

Subscribe

Episode Two: Principle vs. Profit: How Institutions Lose Their Way

Bribery is the corruption we prosecute. But according to Lawrence Lessig, it’s institutional corruption that poses the most danger to American democracy. Hosts Katy Glenn Bass and Madhav Khosla speak with the Harvard Law professor who makes a pointed distinction: the corruption hollowing out American institutions isn’t about bribes, it’s structural—the capture of courts, law firms, corporations, and Congress by economic dependencies pull each away from its founding purpose and leave them vulnerable to political pressure. Yet Lessig finds hope in the lawyers who have refused to bend, in universities that eventually drew a line, and in preparing law students to make professional decisions that align with their values.

Transcript: Ep. 2.

Further Reading

• America, Compromised, Lawrence Lessig (2018)
• State and Local Programs Help Take Big Money Out of Politics, Cinthia Illan-Vazquez and Celina Avalos Jaramillo, Brennan Center for Justice (April 4, 2025)
• Letter to Our Students—An open letter signed by Lawrence Lessig and dozens of Harvard Law School colleagues, addressed to their students, asserting that the United States is experiencing an unprecedented defiance of the rule of law. (3/30/2025)
• Anthropic Goes All-In on Legal, Releasing More Than 20 Connectors and 12 Practice-Area Plugins for Claude, Bob Ambrogi, LawSites (5/12/2026)
• Dinner Table Action v. Schneider—Brennan Center for Justice (2025)

Episode One: What Does Legal Authoritarianism Look Like?

What does authoritarianism look like when it operates through law? In the first episode of “Lawyer Without Law,” hosts Katy Glenn Bass and Madhav Khosla speak with Princeton University Professor Kim Lane Scheppele. They explore historic examples of the legal profession’s role in democratic backsliding around the world and in the United States. They examine how legal systems can consolidate power while maintaining the appearance of legitimacy—and what that means for lawyers and legal institutions as democratic norms come under strain.

Transcript: Ep. 1. 

Further Reading

• At Harvard, Mitt Romney Warns Against ‘Authoritarian’ Presidential Power, by Schuyler Velasco, Harvard Magazine (4/16/2026)
• What Are We Living Through in Trump 2.0? Here Are 3 Possibilities. by Jedediah Britton-Purdy, David Pozen, and John Guida, The New York Times (3/13/2026)
• Kim Lane Scheppele on Hungary, Paul Krugman (Substack) (4/18/2026)
• Collateral Knowledge: Legal Reasoning in the Global Financial Markets, by Fleur Johns and Annelise Riles, European Journal of International Law (12/17/2012) 
• “Ibram X. Kendi and Heather McGhee on the world’s most dangerous conspiracy theory,” @penguinrandomhouse (YouTube) (3/16/2026)
• The professions in America today, by Howard Earl Gardner and Lee S. Shulman, Dædalus (2005)